abs
stringlengths 44
64
| Download PDF
stringlengths 75
115
| OpenReview
stringlengths 42
42
| title
stringlengths 15
148
| url
stringlengths 44
64
| authors
stringlengths 6
903
| detail_url
stringlengths 44
64
| tags
stringclasses 1
value | abstract
stringlengths 422
5.84k
|
|---|---|---|---|---|---|---|---|---|
https://proceedings.mlr.press/v235/cachet24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cachet24a/cachet24a.pdf | https://openreview.net/forum?id=ZrM67ZZ5vj | Bridging Environments and Language with Rendering Functions and Vision-Language Models | https://proceedings.mlr.press/v235/cachet24a.html | Theo Cachet, Christopher R Dance, Olivier Sigaud | https://proceedings.mlr.press/v235/cachet24a.html | ICML 2024 | Vision-language models (VLMs) have tremendous potential for grounding language, and thus enabling language-conditioned agents (LCAs) to perform diverse tasks specified with text. This has motivated the study of LCAs based on reinforcement learning (RL) with rewards given by rendering images of an environment and evaluating those images with VLMs. If single-task RL is employed, such approaches are limited by the cost and time required to train a policy for each new task. Multi-task RL (MTRL) is a natural alternative, but requires a carefully designed corpus of training tasks and does not always generalize reliably to new tasks. Therefore, this paper introduces a novel decomposition of the problem of building an LCA: first find an environment configuration that has a high VLM score for text describing a task; then use a (pretrained) goal-conditioned policy to reach that configuration. We also explore several enhancements to the speed and quality of VLM-based LCAs, notably, the use of distilled models, and the evaluation of configurations from multiple viewpoints to resolve the ambiguities inherent in a single 2D view. We demonstrate our approach on the Humanoid environment, showing that it results in LCAs that outperform MTRL baselines in zero-shot generalization, without requiring any textual task descriptions or other forms of environment-specific annotation during training. |
https://proceedings.mlr.press/v235/cai24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24a/cai24a.pdf | https://openreview.net/forum?id=PnyYgWMMwj | Vocabulary for Universal Approximation: A Linguistic Perspective of Mapping Compositions | https://proceedings.mlr.press/v235/cai24a.html | Yongqiang Cai | https://proceedings.mlr.press/v235/cai24a.html | ICML 2024 | In recent years, deep learning-based sequence modelings, such as language models, have received much attention and success, which pushes researchers to explore the possibility of transforming non-sequential problems into a sequential form. Following this thought, deep neural networks can be represented as composite functions of a sequence of mappings, linear or nonlinear, where each composition can be viewed as a word. However, the weights of linear mappings are undetermined and hence require an infinite number of words. In this article, we investigate the finite case and constructively prove the existence of a finite vocabulary $V$=$\phi_i: \mathbb{R}^d \to \mathbb{R}^d | i=1,...,n$ with $n=O(d^2)$ for the universal approximation. That is, for any continuous mapping $f: \mathbb{R}^d \to \mathbb{R}^d$, compact domain $\Omega$ and $\varepsilon>0$, there is a sequence of mappings $\phi_{i_1}, ..., \phi_{i_m} \in V, m \in \mathbb{Z}^+$, such that the composition $\phi_{i_m} \circ ... \circ \phi_{i_1} $ approximates $f$ on $\Omega$ with an error less than $\varepsilon$. Our results demonstrate an unusual approximation power of mapping compositions and motivate a novel compositional model for regular languages. |
https://proceedings.mlr.press/v235/cai24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24b/cai24b.pdf | https://openreview.net/forum?id=PEpbUobfJv | Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads | https://proceedings.mlr.press/v235/cai24b.html | Tianle Cai, Yuhong Li, Zhengyang Geng, Hongwu Peng, Jason D. Lee, Deming Chen, Tri Dao | https://proceedings.mlr.press/v235/cai24b.html | ICML 2024 | Large Language Models (LLMs) employ auto-regressive decoding that requires sequential computation, with each step reliant on the previous one’s output. This creates a bottleneck as each step necessitates moving the full model parameters from High-Bandwidth Memory (HBM) to the accelerator’s cache. While methods such as speculative decoding have been suggested to address this issue, their implementation is impeded by the challenges associated with acquiring and maintaining a separate draft model. In this paper, we present Medusa, an efficient method that augments LLM inference by adding extra decoding heads to predict multiple subsequent tokens in parallel. Using a tree-based attention mechanism, Medusa constructs multiple candidate continuations and verifies them simultaneously in each decoding step. By leveraging parallel processing, Medusa reduces the number of decoding steps required. We present two levels of fine-tuning procedures for Medusa to meet the needs of different use cases: Medusa-1: Medusa is directly fine-tuned on top of a frozen backbone LLM, enabling lossless inference acceleration. Medusa-2: Medusa is fine-tuned together with the backbone LLM, enabling better prediction accuracy of Medusa heads and higher speedup but needing a special training recipe that preserves the model’s capabilities. Moreover, we propose several extensions that improve or expand the utility of Medusa, including a self-distillation to handle situations where no training data is available and a typical acceptance scheme to boost the acceptance rate while maintaining generation quality. We evaluate Medusa on models of various sizes and training procedures. Our experiments demonstrate that Medusa-1 can achieve over 2.2$\times$ speedup without compromising generation quality, while Medusa-2 further improves the speedup to 2.3-2.8$\times$. |
https://proceedings.mlr.press/v235/cai24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24c/cai24c.pdf | https://openreview.net/forum?id=YlcSyCz21c | Enhancing Cross-Modal Fine-Tuning with Gradually Intermediate Modality Generation | https://proceedings.mlr.press/v235/cai24c.html | Lincan Cai, Shuang Li, Wenxuan Ma, Jingxuan Kang, Binhui Xie, Zixun Sun, Chengwei Zhu | https://proceedings.mlr.press/v235/cai24c.html | ICML 2024 | Large-scale pretrained models have proven immensely valuable in handling data-intensive modalities like text and image. However, fine-tuning these models for certain specialized modalities, such as protein sequence and cosmic ray, poses challenges due to the significant modality discrepancy and scarcity of labeled data. In this paper, we propose an end-to-end method, PaRe, to enhance cross-modal fine-tuning, aiming to transfer a large-scale pretrained model to various target modalities. PaRe employs a gating mechanism to select key patches from both source and target data. Through a modality-agnostic Patch Replacement scheme, these patches are preserved and combined to construct data-rich intermediate modalities ranging from easy to hard. By gradually intermediate modality generation, we can not only effectively bridge the modality gap to enhance stability and transferability of cross-modal fine-tuning, but also address the challenge of limited data in the target modality by leveraging enriched intermediate modality data. Compared with hand-designed, general-purpose, task-specific, and state-of-the-art cross-modal fine-tuning approaches, PaRe demonstrates superior performance across three challenging benchmarks, encompassing more than ten modalities. |
https://proceedings.mlr.press/v235/cai24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24d/cai24d.pdf | https://openreview.net/forum?id=bplNmU2ROC | Batch and match: black-box variational inference with a score-based divergence | https://proceedings.mlr.press/v235/cai24d.html | Diana Cai, Chirag Modi, Loucas Pillaud-Vivien, Charles Margossian, Robert M. Gower, David Blei, Lawrence K. Saul | https://proceedings.mlr.press/v235/cai24d.html | ICML 2024 | Most leading implementations of black-box variational inference (BBVI) are based on optimizing a stochastic evidence lower bound (ELBO). But such approaches to BBVI often converge slowly due to the high variance of their gradient estimates and their sensitivity to hyperparameters. In this work, we propose batch and match (BaM), an alternative approach to BBVI based on a score-based divergence. Notably, this score-based divergence can be optimized by a closed-form proximal update for Gaussian variational families with full covariance matrices. We analyze the convergence of BaM when the target distribution is Gaussian, and we prove that in the limit of infinite batch size the variational parameter updates converge exponentially quickly to the target mean and covariance. We also evaluate the performance of BaM on Gaussian and non-Gaussian target distributions that arise from posterior inference in hierarchical and deep generative models. In these experiments, we find that BaM typically converges in fewer (and sometimes significantly fewer) gradient evaluations than leading implementations of BBVI based on ELBO maximization. |
https://proceedings.mlr.press/v235/cai24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24e/cai24e.pdf | https://openreview.net/forum?id=9vKRhnflAs | Flextron: Many-in-One Flexible Large Language Model | https://proceedings.mlr.press/v235/cai24e.html | Ruisi Cai, Saurav Muralidharan, Greg Heinrich, Hongxu Yin, Zhangyang Wang, Jan Kautz, Pavlo Molchanov | https://proceedings.mlr.press/v235/cai24e.html | ICML 2024 | Training modern LLMs is extremely resource intensive, and customizing them for various deployment scenarios characterized by limited compute and memory resources through repeated training is impractical. In this paper, we introduce Flextron, a network architecture and post-training model optimization framework supporting flexible model deployment. The Flextron architecture utilizes a nested elastic structure to rapidly adapt to specific user-defined latency and accuracy targets during inference with no additional fine-tuning required. It is also input-adaptive, and can automatically route tokens through its sub-networks for improved performance and efficiency. We present a sample-efficient training method and associated routing algorithms for systematically transforming an existing trained LLM into a Flextron model. We evaluate Flextron on the GPT-3 and LLama-2 family of LLMs, and demonstrate superior performance over multiple end-to-end trained variants and other state-of-the-art elastic networks, all with a single pretraining run that consumes a mere 7.63% tokens compared to original pretraining. |
https://proceedings.mlr.press/v235/cai24f.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24f/cai24f.pdf | https://openreview.net/forum?id=EK7fuAMNoI | Accelerated Algorithms for Constrained Nonconvex-Nonconcave Min-Max Optimization and Comonotone Inclusion | https://proceedings.mlr.press/v235/cai24f.html | Yang Cai, Argyris Oikonomou, Weiqiang Zheng | https://proceedings.mlr.press/v235/cai24f.html | ICML 2024 | We study constrained comonotone min-max optimization, a structured class of nonconvex-nonconcave min-max optimization problems, and their generalization to comonotone inclusion. In our first contribution, we extend the Extra Anchored Gradient (EAG) algorithm, originally proposed by Yoon and Ryu (2021) for unconstrained min-max optimization, to constrained comonotone min-max optimization and comonotone inclusion, achieving an optimal convergence rate of $O\left(\frac{1}{T}\right)$ among all first-order methods. Additionally, we prove that the algorithm’s iterations converge to a point in the solution set. In our second contribution, we extend the Fast Extra Gradient (FEG) algorithm, as developed by Lee and Kim (2021), to constrained comonotone min-max optimization and comonotone inclusion, achieving the same $O\left(\frac{1}{T}\right)$ convergence rate. This rate is applicable to the broadest set of comonotone inclusion problems yet studied in the literature. Our analyses are based on simple potential function arguments, which might be useful for analyzing other accelerated algorithms. |
https://proceedings.mlr.press/v235/cai24g.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24g/cai24g.pdf | https://openreview.net/forum?id=NUlyqMyhO9 | LoCoCo: Dropping In Convolutions for Long Context Compression | https://proceedings.mlr.press/v235/cai24g.html | Ruisi Cai, Yuandong Tian, Zhangyang Wang, Beidi Chen | https://proceedings.mlr.press/v235/cai24g.html | ICML 2024 | This paper tackles the memory hurdle of of processing long context sequences in Large Language Models (LLMs), by presenting a novel approach, Dropping In Convolutions for Long Context Compression (LoCoCo). LoCoCo employs only a fixed-size Key-Value (KV) cache, and can enhance efficiency in both inference and fine-tuning stages. Diverging from prior methods that selectively drop KV pairs based on heuristics, LoCoCo leverages a data-driven adaptive fusion technique, blending previous KV pairs with incoming tokens to minimize the loss of contextual information and ensure accurate attention modeling. This token integration is achieved through injecting one-dimensional convolutional kernels that dynamically calculate mixing weights for each KV cache slot. Designed for broad compatibility with existing LLM frameworks, LoCoCo allows for straightforward "drop-in" integration without needing architectural modifications, while incurring minimal tuning overhead. Experiments demonstrate that LoCoCo maintains consistently outstanding performance across various context lengths and can achieve a high context compression rate during both inference and fine-tuning phases. During inference, we successfully compressed up to $3482$ tokens into a $128$-size KV cache, while retaining comparable performance to the full sequence - an accuracy improvement of up to $0.2791$ compared to baselines at the same cache size. During post-training tuning, we also effectively extended the context length from 4K to 32K using a KV cache of fixed size 512, achieving performance similar to fine-tuning with entire sequences. |
https://proceedings.mlr.press/v235/cai24h.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24h/cai24h.pdf | https://openreview.net/forum?id=YB1O99gK7b | On Gradient-like Explanation under a Black-box Setting: When Black-box Explanations Become as Good as White-box | https://proceedings.mlr.press/v235/cai24h.html | Yi Cai, Gerhard Wunder | https://proceedings.mlr.press/v235/cai24h.html | ICML 2024 | Attribution methods shed light on the explainability of data-driven approaches such as deep learning models by uncovering the most influential features in a to-be-explained decision. While determining feature attributions via gradients delivers promising results, the internal access required for acquiring gradients can be impractical under safety concerns, thus limiting the applicability of gradient-based approaches. In response to such limited flexibility, this paper presents GEEX (gradient-estimation-based explanation), a method that produces gradient-like explanations through only query-level access. The proposed approach holds a set of fundamental properties for attribution methods, which are mathematically rigorously proved, ensuring the quality of its explanations. In addition to the theoretical analysis, with a focus on image data, the experimental results empirically demonstrate the superiority of the proposed method over state-of-the-art black-box methods and its competitive performance compared to methods with full access. |
https://proceedings.mlr.press/v235/cai24i.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cai24i/cai24i.pdf | https://openreview.net/forum?id=4sikyurTLX | Sample-specific Masks for Visual Reprogramming-based Prompting | https://proceedings.mlr.press/v235/cai24i.html | Chengyi Cai, Zesheng Ye, Lei Feng, Jianzhong Qi, Feng Liu | https://proceedings.mlr.press/v235/cai24i.html | ICML 2024 | Visual reprogramming (VR) is a prompting technique that aims to re-purpose a pre-trained model (e.g., a classifier on ImageNet) to target tasks (e.g., medical data prediction) by learning a small-scale pattern added into input images instead of tuning considerable parameters within the model. The location of the pattern within input samples is usually determined by a pre-defined mask shared across all samples. In this paper, we show that the shared mask potentially limits VR’s generalization and increases its approximation error due to the lack of sample-level adaptation. Motivated by this finding, we design a new framework for VR called sample-specific multi-channel masks (SMM). Specifically, SMM employs a lightweight ConvNet and patch-wise interpolation to generate sample-specific three-channel masks instead of a shared and pre-defined mask. Since we generate different masks for individual samples, SMM is theoretically shown to reduce approximation error for the target tasks compared with existing state-of-the-art VR methods. We also empirically demonstrate its performance gain on both ResNet and ViT. The success of SMM further highlights the broader applicability of VR in leveraging the latent knowledge of pre-trained models for various target tasks. Our code is available at https://github.com/tmlr-group/SMM. |
https://proceedings.mlr.press/v235/calandriello24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/calandriello24a/calandriello24a.pdf | https://openreview.net/forum?id=2RQqg2Y7Y6 | Human Alignment of Large Language Models through Online Preference Optimisation | https://proceedings.mlr.press/v235/calandriello24a.html | Daniele Calandriello, Zhaohan Daniel Guo, Remi Munos, Mark Rowland, Yunhao Tang, Bernardo Avila Pires, Pierre Harvey Richemond, Charline Le Lan, Michal Valko, Tianqi Liu, Rishabh Joshi, Zeyu Zheng, Bilal Piot | https://proceedings.mlr.press/v235/calandriello24a.html | ICML 2024 | Ensuring alignment of language model’s outputs with human preferences is critical to guarantee a useful, safe, and pleasant user experience. Thus, human alignment has been extensively studied recently and several methods such as Reinforcement Learning from Human Feedback (RLHF), Direct Policy Optimisation (DPO) and Sequence Likelihood Calibration (SLiC) have emerged. In this paper, our contribution is two-fold. First, we show the equivalence between two recent alignment methods, namely Identity Policy Optimisation (IPO) and Nash Mirror Descent (Nash-MD). Second, we introduce a generalisation of IPO, named IPO-MD, that leverages the regularised sampling approach proposed by Nash-MD. This equivalence may seem surprising at first sight, since IPO is an offline method whereas Nash-MD is an online method using a preference model. However, this equivalence can be proven when we consider the online version of IPO, that is when both generations are sampled by the online policy and annotated by a trained preference model. Optimising the IPO loss with such a stream of data becomes then equivalent to finding the Nash equilibrium of the preference model through self-play. Building on this equivalence, we introduce the IPO-MD algorithm that generates data with a mixture policy (between the online and reference policy) similarly as the general Nash-MD algorithm. We compare online-IPO and IPO-MD to different online versions of existing losses on preference data such as DPO and SLiC on a summarisation task. |
https://proceedings.mlr.press/v235/calvo-ordonez24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/calvo-ordonez24a/calvo-ordonez24a.pdf | https://openreview.net/forum?id=jNab9mXEyj | Partially Stochastic Infinitely Deep Bayesian Neural Networks | https://proceedings.mlr.press/v235/calvo-ordonez24a.html | Sergio Calvo Ordoñez, Matthieu Meunier, Francesco Piatti, Yuantao Shi | https://proceedings.mlr.press/v235/calvo-ordonez24a.html | ICML 2024 | In this paper, we present Partially Stochastic Infinitely Deep Bayesian Neural Networks, a novel family of architectures that integrates partial stochasticity into the framework of infinitely deep neural networks. Our new class of architectures is designed to improve the computational efficiency of existing architectures at training and inference time. To do this, we leverage the advantages of partial stochasticity in the infinite-depth limit which include the benefits of full stochasticity e.g. robustness, uncertainty quantification, and memory efficiency, whilst improving their limitations around computational complexity. We present a variety of architectural configurations, offering flexibility in network design including different methods for weight partition. We also provide mathematical guarantees on the expressivity of our models by establishing that our network family qualifies as Universal Conditional Distribution Approximators. Lastly, empirical evaluations across multiple tasks show that our proposed architectures achieve better downstream task performance and uncertainty quantification than their counterparts while being significantly more efficient. The code can be found at https://github.com/Sergio20f/part_stoch_inf_deep |
https://proceedings.mlr.press/v235/campbell24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/campbell24a/campbell24a.pdf | https://openreview.net/forum?id=kQwSbv0BR4 | Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design | https://proceedings.mlr.press/v235/campbell24a.html | Andrew Campbell, Jason Yim, Regina Barzilay, Tom Rainforth, Tommi Jaakkola | https://proceedings.mlr.press/v235/campbell24a.html | ICML 2024 | Combining discrete and continuous data is an important capability for generative models. We present Discrete Flow Models (DFMs), a new flow-based model of discrete data that provides the missing link in enabling flow-based generative models to be applied to multimodal continuous and discrete data problems. Our key insight is that the discrete equivalent of continuous space flow matching can be realized using Continuous Time Markov Chains. DFMs benefit from a simple derivation that includes discrete diffusion models as a specific instance while allowing improved performance over existing diffusion-based approaches. We utilize our DFMs method to build a multimodal flow-based modeling framework. We apply this capability to the task of protein co-design, wherein we learn a model for jointly generating protein structure and sequence. Our approach achieves state-of-the-art co-design performance while allowing the same multimodal model to be used for flexible generation of the sequence or structure. |
https://proceedings.mlr.press/v235/candido-ramos24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/candido-ramos24a/candido-ramos24a.pdf | https://openreview.net/forum?id=JAfIDm7NED | Mimicking Better by Matching the Approximate Action Distribution | https://proceedings.mlr.press/v235/candido-ramos24a.html | Joao Candido Ramos, Lionel Blondé, Naoya Takeishi, Alexandros Kalousis | https://proceedings.mlr.press/v235/candido-ramos24a.html | ICML 2024 | In this paper, we introduce MAAD, a novel, sample-efficient on-policy algorithm for Imitation Learning from Observations. MAAD utilizes a surrogate reward signal, which can be derived from various sources such as adversarial games, trajectory matching objectives, or optimal transport criteria. To compensate for the non-availability of expert actions, we rely on an inverse dynamics model that infers plausible actions distribution given the expert’s state-state transitions; we regularize the imitator’s policy by aligning it to the inferred action distribution. MAAD leads to significantly improved sample efficiency and stability. We demonstrate its effectiveness in a number of MuJoCo environments, both int the OpenAI Gym and the DeepMind Control Suite. We show that it requires considerable fewer interactions to achieve expert performance, outperforming current state-of-the-art on-policy methods. Remarkably, MAAD often stands out as the sole method capable of attaining expert performance levels, underscoring its simplicity and efficacy. |
https://proceedings.mlr.press/v235/canturk24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/canturk24a/canturk24a.pdf | https://openreview.net/forum?id=UTSCK582Yo | Graph Positional and Structural Encoder | https://proceedings.mlr.press/v235/canturk24a.html | Semih Cantürk, Renming Liu, Olivier Lapointe-Gagné, Vincent Létourneau, Guy Wolf, Dominique Beaini, Ladislav Rampášek | https://proceedings.mlr.press/v235/canturk24a.html | ICML 2024 | Positional and structural encodings (PSE) enable better identifiability of nodes within a graph, rendering them essential tools for empowering modern GNNs, and in particular graph Transformers. However, designing PSEs that work optimally for all graph prediction tasks is a challenging and unsolved problem. Here, we present the Graph Positional and Structural Encoder (GPSE), the first-ever graph encoder designed to capture rich PSE representations for augmenting any GNN. GPSE learns an efficient common latent representation for multiple PSEs, and is highly transferable: The encoder trained on a particular graph dataset can be used effectively on datasets drawn from markedly different distributions and modalities. We show that across a wide range of benchmarks, GPSE-enhanced models can significantly outperform those that employ explicitly computed PSEs, and at least match their performance in others. Our results pave the way for the development of foundational pre-trained graph encoders for extracting positional and structural information, and highlight their potential as a more powerful and efficient alternative to explicitly computed PSEs and existing self-supervised pre-training approaches. Our framework and pre-trained models are publicly available at https://github.com/G-Taxonomy-Workgroup/GPSE. For convenience, GPSE has also been integrated into the PyG library to facilitate downstream applications. |
https://proceedings.mlr.press/v235/cao24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cao24a/cao24a.pdf | https://openreview.net/forum?id=LJcIIhqGDN | Successor Features for Efficient Multi-Subject Controlled Text Generation | https://proceedings.mlr.press/v235/cao24a.html | Meng Cao, Mehdi Fatemi, Jackie Ck Cheung, Samira Shabanian | https://proceedings.mlr.press/v235/cao24a.html | ICML 2024 | While large language models (LLMs) have achieved impressive performance in generating fluent and realistic text, controlling the generated text so that it exhibits properties such as safety, factuality, and non-toxicity remains challenging. Existing decoding-based controllable text generation methods are static in terms of the dimension of control; if the target subject is changed, they require new training. Moreover, it can quickly become prohibitive to concurrently control multiple subjects. To address these challenges, we first show that existing methods can be framed as a reinforcement learning problem, where an action-value function estimates the likelihood of a desired attribute appearing in the generated text. Then, we introduce a novel approach named SF-Gen, which leverages the concept of successor features to decouple the dynamics of LLMs from task-specific rewards. By employing successor features, our method proves to be memory-efficient and computationally efficient for both training and decoding, especially when dealing with multiple target subjects. To the best of our knowledge, our research represents the first application of successor features in text generation. In addition to its computational efficiency, the resultant language produced by our method is comparable to the SOTA (and outperforms baselines) in both control measures as well as language quality, which we demonstrate through a series of experiments in various controllable text generation tasks. |
https://proceedings.mlr.press/v235/cao24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cao24b/cao24b.pdf | https://openreview.net/forum?id=PAbkWU0KDG | Limited Preference Aided Imitation Learning from Imperfect Demonstrations | https://proceedings.mlr.press/v235/cao24b.html | Xingchen Cao, Fan-Ming Luo, Junyin Ye, Tian Xu, Zhilong Zhang, Yang Yu | https://proceedings.mlr.press/v235/cao24b.html | ICML 2024 | Imitation learning mimics high-quality policies from expert data for sequential decision-making tasks. However, its efficacy is hindered in scenarios where optimal demonstrations are unavailable, and only imperfect demonstrations are present. To address this issue, introducing additional limited human preferences is a suitable approach as it can be obtained in a human-friendly manner, offering a promising way to learn the policy that exceeds the performance of imperfect demonstrations. In this paper, we propose a novel imitation learning (IL) algorithm, Preference Aided Imitation Learning from imperfect demonstrations (PAIL). Specifically, PAIL learns a preference reward by querying experts for limited preferences from imperfect demonstrations. This serves two purposes during training: 1) Reweighting imperfect demonstrations with the preference reward for higher quality. 2) Selecting explored trajectories with high cumulative preference rewards to augment imperfect demonstrations. The dataset with continuously improving quality empowers the performance of PAIL to transcend the initial demonstrations. Comprehensive empirical results across a synthetic task and two locomotion benchmarks show that PAIL surpasses baselines by 73.2% and breaks through the performance bottleneck of imperfect demonstrations. |
https://proceedings.mlr.press/v235/cao24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cao24c/cao24c.pdf | https://openreview.net/forum?id=LYpGLrC4oq | Predictive Dynamic Fusion | https://proceedings.mlr.press/v235/cao24c.html | Bing Cao, Yinan Xia, Yi Ding, Changqing Zhang, Qinghua Hu | https://proceedings.mlr.press/v235/cao24c.html | ICML 2024 | Multimodal fusion is crucial in joint decision-making systems for rendering holistic judgments. Since multimodal data changes in open environments, dynamic fusion has emerged and achieved remarkable progress in numerous applications. However, most existing dynamic multimodal fusion methods lack theoretical guarantees and easily fall into suboptimal problems, yielding unreliability and instability. To address this issue, we propose a Predictive Dynamic Fusion (PDF) framework for multimodal learning. We proceed to reveal the multimodal fusion from a generalization perspective and theoretically derive the predictable Collaborative Belief (Co-Belief) with Mono- and Holo-Confidence, which provably reduces the upper bound of generalization error. Accordingly, we further propose a relative calibration strategy to calibrate the predicted Co-Belief for potential uncertainty. Extensive experiments on multiple benchmarks confirm our superiority. Our code is available at https://github.com/Yinan-Xia/PDF. |
https://proceedings.mlr.press/v235/cao24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cao24d/cao24d.pdf | https://openreview.net/forum?id=xZO7SmM12y | Envisioning Outlier Exposure by Large Language Models for Out-of-Distribution Detection | https://proceedings.mlr.press/v235/cao24d.html | Chentao Cao, Zhun Zhong, Zhanke Zhou, Yang Liu, Tongliang Liu, Bo Han | https://proceedings.mlr.press/v235/cao24d.html | ICML 2024 | Detecting out-of-distribution (OOD) samples is essential when deploying machine learning models in open-world scenarios. Zero-shot OOD detection, requiring no training on in-distribution (ID) data, has been possible with the advent of vision-language models like CLIP. Existing methods build a text-based classifier with only closed-set labels. However, this largely restricts the inherent capability of CLIP to recognize samples from large and open label space. In this paper, we propose to tackle this constraint by leveraging the expert knowledge and reasoning capability of large language models (LLM) to Envision potential Outlier Exposure, termed EOE, without access to any actual OOD data. Owing to better adaptation to open-world scenarios, EOE can be generalized to different tasks, including far, near, and fine-grained OOD detection. Technically, we design (1) LLM prompts based on visual similarity to generate potential outlier class labels specialized for OOD detection, as well as (2) a new score function based on potential outlier penalty to distinguish hard OOD samples effectively. Empirically, EOE achieves state-of-the-art performance across different OOD tasks and can be effectively scaled to the ImageNet-1K dataset. The code is publicly available at: https://github.com/tmlr-group/EOE. |
https://proceedings.mlr.press/v235/caragiannis24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/caragiannis24a/caragiannis24a.pdf | https://openreview.net/forum?id=nsjfoziR5j | Can a Few Decide for Many? The Metric Distortion of Sortition | https://proceedings.mlr.press/v235/caragiannis24a.html | Ioannis Caragiannis, Evi Micha, Jannik Peters | https://proceedings.mlr.press/v235/caragiannis24a.html | ICML 2024 | Recent works have studied the design of algorithms for selecting representative sortition panels. However, the most central question remains unaddressed: Do these panels reflect the entire population’s opinion? We present a positive answer by adopting the concept of metric distortion from computational social choice, which aims to quantify how much a panel’s decision aligns with the ideal decision of the population when preferences and agents lie on a metric space. We show that uniform selection needs only logarithmically many agents in terms of the number of alternatives to achieve almost optimal distortion. We also show that Fair Greedy Capture, a selection algorithm introduced recently by Ebadian and Micha (2024), matches uniform selection’s guarantees of almost optimal distortion and also achieves constant ex-post distortion, ensuring a “best of both worlds” performance. |
https://proceedings.mlr.press/v235/carlini24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/carlini24a/carlini24a.pdf | https://openreview.net/forum?id=VE3yWXt3KB | Stealing part of a production language model | https://proceedings.mlr.press/v235/carlini24a.html | Nicholas Carlini, Daniel Paleka, Krishnamurthy Dj Dvijotham, Thomas Steinke, Jonathan Hayase, A. Feder Cooper, Katherine Lee, Matthew Jagielski, Milad Nasr, Arthur Conmy, Eric Wallace, David Rolnick, Florian Tramèr | https://proceedings.mlr.press/v235/carlini24a.html | ICML 2024 | We introduce the first model-stealing attack that extracts precise, nontrivial information from black-box production language models like OpenAI’s ChatGPT or Google’s PaLM-2. Specifically, our attack recovers the embedding projection layer (up to symmetries) of a transformer model, given typical API access. For under $20 USD, our attack extracts the entire projection matrix of OpenAI’s Ada and Babbage language models. We thereby confirm, for the first time, that these black-box models have a hidden dimension of 1024 and 2048, respectively. We also recover the exact hidden dimension size of the GPT-3.5-turbo model, and estimate it would cost under \$2,000 in queries to recover the entire projection matrix. We conclude with potential defenses and mitigations, and discuss the implications of possible future work that could extend our attack. |
https://proceedings.mlr.press/v235/carroll24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/carroll24a/carroll24a.pdf | https://openreview.net/forum?id=itYGbe0Cs1 | AI Alignment with Changing and Influenceable Reward Functions | https://proceedings.mlr.press/v235/carroll24a.html | Micah Carroll, Davis Foote, Anand Siththaranjan, Stuart Russell, Anca Dragan | https://proceedings.mlr.press/v235/carroll24a.html | ICML 2024 | Existing AI alignment approaches assume that preferences are static, which is unrealistic: our preferences change, and may even be influenced by our interactions with AI systems themselves. To clarify the consequences of incorrectly assuming static preferences, we introduce Dynamic Reward Markov Decision Processes (DR-MDPs), which explicitly model preference changes and the AI’s influence on them. We show that despite its convenience, the static-preference assumption may undermine the soundness of existing alignment techniques, leading them to implicitly reward AI systems for influencing user preferences in ways users may not truly want. We then explore potential solutions. First, we offer a unifying perspective on how an agent’s optimization horizon may partially help reduce undesirable AI influence. Then, we formalize different notions of AI alignment that account for preference change from the outset. Comparing the strengths and limitations of 8 such notions of alignment, we find that they all either err towards causing undesirable AI influence, or are overly risk-averse, suggesting that a straightforward solution to the problems of changing preferences may not exist. As there is no avoiding grappling with changing preferences in real-world settings, this makes it all the more important to handle these issues with care, balancing risks and capabilities. We hope our work can provide conceptual clarity and constitute a first step towards AI alignment practices which explicitly account for (and contend with) the changing and influenceable nature of human preferences. |
https://proceedings.mlr.press/v235/cassel24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cassel24a/cassel24a.pdf | https://openreview.net/forum?id=hXQOO6VsxH | Near-Optimal Regret in Linear MDPs with Aggregate Bandit Feedback | https://proceedings.mlr.press/v235/cassel24a.html | Asaf Cassel, Haipeng Luo, Aviv Rosenberg, Dmitry Sotnikov | https://proceedings.mlr.press/v235/cassel24a.html | ICML 2024 | In many real-world applications, it is hard to provide a reward signal in each step of a Reinforcement Learning (RL) process and more natural to give feedback when an episode ends. To this end, we study the recently proposed model of RL with Aggregate Bandit Feedback (RL-ABF), where the agent only observes the sum of rewards at the end of an episode instead of each reward individually. Prior work studied RL-ABF only in tabular settings, where the number of states is assumed to be small. In this paper, we extend ABF to linear function approximation and develop two efficient algorithms with near-optimal regret guarantees: a value-based optimistic algorithm built on a new randomization technique with a Q-functions ensemble, and a policy optimization algorithm that uses a novel hedging scheme over the ensemble. |
https://proceedings.mlr.press/v235/castiglioni24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/castiglioni24a/castiglioni24a.pdf | https://openreview.net/forum?id=shzEkKPrsn | Online Learning under Budget and ROI Constraints via Weak Adaptivity | https://proceedings.mlr.press/v235/castiglioni24a.html | Matteo Castiglioni, Andrea Celli, Christian Kroer | https://proceedings.mlr.press/v235/castiglioni24a.html | ICML 2024 | We study online learning problems in which a decision maker has to make a sequence of costly decisions, with the goal of maximizing their expected reward while adhering to budget and return-on-investment (ROI) constraints. Existing primal-dual algorithms designed for constrained online learning problems under adversarial inputs rely on two fundamental assumptions. First, the decision maker must know beforehand the value of parameters related to the degree of strict feasibility of the problem (i.e. Slater parameters). Second, a strictly feasible solution to the offline optimization problem must exist at each round. Both requirements are unrealistic for practical applications such as bidding in online ad auctions. In this paper, we show how such assumptions can be circumvented by endowing standard primal-dual templates with weakly adaptive regret minimizers. This results in a “dual-balancing” framework which ensures that dual variables stay sufficiently small, even in the absence of knowledge about Slater’s parameter. We prove the first best-of-both-worlds no-regret guarantees which hold in absence of the two aforementioned assumptions, under stochastic and adversarial inputs. Finally, we show how to instantiate the framework to optimally bid in various mechanisms of practical relevance, such as first- and second-price auctions. |
https://proceedings.mlr.press/v235/castin24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/castin24a/castin24a.pdf | https://openreview.net/forum?id=aP0H8A1ywk | How Smooth Is Attention? | https://proceedings.mlr.press/v235/castin24a.html | Valérie Castin, Pierre Ablin, Gabriel Peyré | https://proceedings.mlr.press/v235/castin24a.html | ICML 2024 | Self-attention and masked self-attention are at the heart of Transformers’ outstanding success. Still, our mathematical understanding of attention, in particular of its Lipschitz properties — which are key when it comes to analyzing robustness and expressive power — is incomplete. We provide a detailed study of the Lipschitz constant of self-attention in several practical scenarios, discussing the impact of the sequence length $n$ and layer normalization on the local Lipschitz constant of both unmasked and masked self-attention. In particular, we show that for inputs of length $n$ in any compact set, the Lipschitz constant of self-attention is bounded by $\sqrt{n}$ up to a constant factor and that this bound is tight for reasonable sequence lengths. When the sequence length $n$ is too large for the previous bound to be tight, which we refer to as the mean-field regime, we provide an upper bound and a matching lower bound which are independent of $n$. Our mean-field framework for masked self-attention is novel and of independent interest. Our experiments on pretrained and randomly initialized BERT and GPT-2 support our theoretical findings. |
https://proceedings.mlr.press/v235/catalano24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/catalano24a/catalano24a.pdf | https://openreview.net/forum?id=cmy38XZlJu | Hierarchical Integral Probability Metrics: A distance on random probability measures with low sample complexity | https://proceedings.mlr.press/v235/catalano24a.html | Marta Catalano, Hugo Lavenant | https://proceedings.mlr.press/v235/catalano24a.html | ICML 2024 | Random probabilities are a key component to many nonparametric methods in Statistics and Machine Learning. To quantify comparisons between different laws of random probabilities several works are starting to use the elegant Wasserstein over Wasserstein distance. In this paper we prove that the infinite dimensionality of the space of probabilities drastically deteriorates its sample complexity, which is slower than any polynomial rate in the sample size. We propose a new distance that preserves many desirable properties of the former while achieving a parametric rate of convergence. In particular, our distance 1) metrizes weak convergence; 2) can be estimated numerically through samples with low complexity; 3) can be bounded analytically from above and below. The main ingredient are integral probability metrics, which lead to the name hierarchical IPM. |
https://proceedings.mlr.press/v235/cattaneo24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cattaneo24a/cattaneo24a.pdf | https://openreview.net/forum?id=y8YovS0lOg | On the Implicit Bias of Adam | https://proceedings.mlr.press/v235/cattaneo24a.html | Matias D. Cattaneo, Jason Matthew Klusowski, Boris Shigida | https://proceedings.mlr.press/v235/cattaneo24a.html | ICML 2024 | In previous literature, backward error analysis was used to find ordinary differential equations (ODEs) approximating the gradient descent trajectory. It was found that finite step sizes implicitly regularize solutions because terms appearing in the ODEs penalize the two-norm of the loss gradients. We prove that the existence of similar implicit regularization in RMSProp and Adam depends on their hyperparameters and the training stage, but with a different "norm" involved: the corresponding ODE terms either penalize the (perturbed) one-norm of the loss gradients or, conversely, impede its reduction (the latter case being typical). We also conduct numerical experiments and discuss how the proven facts can influence generalization. |
https://proceedings.mlr.press/v235/celik24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/celik24a/celik24a.pdf | https://openreview.net/forum?id=9ZkUFSwlUH | Acquiring Diverse Skills using Curriculum Reinforcement Learning with Mixture of Experts | https://proceedings.mlr.press/v235/celik24a.html | Onur Celik, Aleksandar Taranovic, Gerhard Neumann | https://proceedings.mlr.press/v235/celik24a.html | ICML 2024 | Reinforcement learning (RL) is a powerful approach for acquiring a good-performing policy. However, learning diverse skills is challenging in RL due to the commonly used Gaussian policy parameterization. We propose Diverse Skill Learning (Di-SkilL), an RL method for learning diverse skills using Mixture of Experts, where each expert formalizes a skill as a contextual motion primitive. Di-SkilL optimizes each expert and its associate context distribution to a maximum entropy objective that incentivizes learning diverse skills in similar contexts. The per-expert context distribution enables automatic curricula learning, allowing each expert to focus on its best-performing sub-region of the context space. To overcome hard discontinuities and multi-modalities without any prior knowledge of the environment’s unknown context probability space, we leverage energy-based models to represent the per-expert context distributions and demonstrate how we can efficiently train them using the standard policy gradient objective. We show on challenging robot simulation tasks that Di-SkilL can learn diverse and performant skills. |
https://proceedings.mlr.press/v235/celis24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/celis24a/celis24a.pdf | https://openreview.net/forum?id=9QRcp2ubDt | Centralized Selection with Preferences in the Presence of Biases | https://proceedings.mlr.press/v235/celis24a.html | L. Elisa Celis, Amit Kumar, Nisheeth K. Vishnoi, Andrew Xu | https://proceedings.mlr.press/v235/celis24a.html | ICML 2024 | This paper considers the scenario in which there are multiple institutions, each with a limited capacity for candidates, and candidates, each with preferences over the institutions. A central entity evaluates the utility of each candidate to the institutions, and the goal is to select candidates for each institution in a way that maximizes utility while also considering the candidates’ preferences. The paper focuses on the setting in which candidates are divided into multiple groups and the observed utilities of candidates in some groups are biased–systematically lower than their true utilities. The first result is that, in these biased settings, prior algorithms can lead to selections with sub-optimal true utility and significant discrepancies in the fraction of candidates from each group that get their preferred choices. Subsequently, an algorithm is presented along with proof that it produces selections that achieve near-optimal group fairness with respect to preferences while also nearly maximizing the true utility under distributional assumptions. Further, extensive empirical validation of these results in real-world and synthetic settings, in which the distributional assumptions may not hold, are presented. |
https://proceedings.mlr.press/v235/cen24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cen24a/cen24a.pdf | https://openreview.net/forum?id=o1gS6MNAw8 | Using Left and Right Brains Together: Towards Vision and Language Planning | https://proceedings.mlr.press/v235/cen24a.html | Jun Cen, Chenfei Wu, Xiao Liu, Shengming Yin, Yixuan Pei, Jinglong Yang, Qifeng Chen, Nan Duan, Jianguo Zhang | https://proceedings.mlr.press/v235/cen24a.html | ICML 2024 | Large Language Models (LLMs) and Large Multi-modality Models (LMMs) have demonstrated remarkable decision masking capabilities on a variety of tasks. However, they inherently operate planning within the language space, lacking the vision and spatial imagination ability. In contrast, humans utilize both left and right hemispheres of the brain for language and visual planning during the thinking process. Therefore, we introduce a novel vision-language planning framework in this work to perform concurrent visual and language planning for tasks with inputs of any form. Our framework incorporates visual planning to capture intricate environmental details, while language planning enhances the logical coherence of the overall system. We evaluate the effectiveness of our framework across vision-language tasks, vision-only tasks, and language-only tasks. The results demonstrate the superior performance of our approach, indicating that the integration of visual and language planning yields better contextually aware task execution. |
https://proceedings.mlr.press/v235/cen24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cen24b/cen24b.pdf | https://openreview.net/forum?id=JNHK11bAGl | Feasibility Consistent Representation Learning for Safe Reinforcement Learning | https://proceedings.mlr.press/v235/cen24b.html | Zhepeng Cen, Yihang Yao, Zuxin Liu, Ding Zhao | https://proceedings.mlr.press/v235/cen24b.html | ICML 2024 | In the field of safe reinforcement learning (RL), finding a balance between satisfying safety constraints and optimizing reward performance presents a significant challenge. A key obstacle in this endeavor is the estimation of safety constraints, which is typically more difficult than estimating a reward metric due to the sparse nature of the constraint signals. To address this issue, we introduce a novel framework named Feasibility Consistent Safe Reinforcement Learning (FCSRL). This framework combines representation learning with feasibility-oriented objectives to identify and extract safety-related information from the raw state for safe RL. Leveraging self-supervised learning techniques and a more learnable safety metric, our approach enhances the policy learning and constraint estimation. Empirical evaluations across a range of vector-state and image-based tasks demonstrate that our method is capable of learning a better safety-aware embedding and achieving superior performance than previous representation learning baselines. |
https://proceedings.mlr.press/v235/cetin24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cetin24a/cetin24a.pdf | https://openreview.net/forum?id=japBn31gXC | Simple Ingredients for Offline Reinforcement Learning | https://proceedings.mlr.press/v235/cetin24a.html | Edoardo Cetin, Andrea Tirinzoni, Matteo Pirotta, Alessandro Lazaric, Yann Ollivier, Ahmed Touati | https://proceedings.mlr.press/v235/cetin24a.html | ICML 2024 | Offline reinforcement learning algorithms have proven effective on datasets highly connected to the target downstream task. Yet, by leveraging a novel testbed (MOOD) in which trajectories come from heterogeneous sources, we show that existing methods struggle with diverse data: their performance considerably deteriorates as data collected for related but different tasks is simply added to the offline buffer. In light of this finding, we conduct a large empirical study where we formulate and test several hypotheses to explain this failure. Surprisingly, we find that targeted scale, more than algorithmic considerations, is the key factor influencing performance. We show that simple methods like AWAC and IQL with increased policy size overcome the paradoxical failure modes from the inclusion of additional data in MOOD, and notably outperform prior state-of-the-art algorithms on the canonical D4RL benchmark. |
https://proceedings.mlr.press/v235/cha24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cha24a/cha24a.pdf | https://openreview.net/forum?id=9jXS07TIBH | Regularizing with Pseudo-Negatives for Continual Self-Supervised Learning | https://proceedings.mlr.press/v235/cha24a.html | Sungmin Cha, Kyunghyun Cho, Taesup Moon | https://proceedings.mlr.press/v235/cha24a.html | ICML 2024 | We introduce a novel Pseudo-Negative Regularization (PNR) framework for effective continual self-supervised learning (CSSL). Our PNR leverages pseudo-negatives obtained through model-based augmentation in a way that newly learned representations may not contradict what has been learned in the past. Specifically, for the InfoNCE-based contrastive learning methods, we define symmetric pseudo-negatives obtained from current and previous models and use them in both main and regularization loss terms. Furthermore, we extend this idea to non-contrastive learning methods which do not inherently rely on negatives. For these methods, a pseudo-negative is defined as the output from the previous model for a differently augmented version of the anchor sample and is asymmetrically applied to the regularization term. Extensive experimental results demonstrate that our PNR framework achieves state-of-the-art performance in representation learning during CSSL by effectively balancing the trade-off between plasticity and stability. |
https://proceedings.mlr.press/v235/chadha24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chadha24a/chadha24a.pdf | https://openreview.net/forum?id=FVmqX0sYz9 | Auditing Private Prediction | https://proceedings.mlr.press/v235/chadha24a.html | Karan Chadha, Matthew Jagielski, Nicolas Papernot, Christopher A. Choquette-Choo, Milad Nasr | https://proceedings.mlr.press/v235/chadha24a.html | ICML 2024 | Differential privacy (DP) offers a theoretical upper bound on the potential privacy leakage of an algorithm, while empirical auditing establishes a practical lower bound. Auditing techniques exist for DP training algorithms. However machine learning can also be made private at inference. We propose the first framework for auditing private prediction where we instantiate adversaries with varying poisoning and query capabilities. This enables us to study the privacy leakage of four private prediction algorithms: PATE (Papernot et al., 2016), CaPC (Choquette-Choo et al., 2020), PromptPATE (Duan et al., 2023), and Private-kNN (Zhu et al., 2020). To conduct our audit, we introduce novel techniques to empirically evaluate privacy leakage in terms of Renyi DP. Our experiments show that (i) the privacy analysis of private prediction can be improved, (ii) algorithms which are easier to poison lead to much higher privacy leakage, and (iii) the privacy leakage is significantly lower for adversaries without query control than those with full control. |
https://proceedings.mlr.press/v235/chakraborty24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chakraborty24a/chakraborty24a.pdf | https://openreview.net/forum?id=CJbhtpcyGL | Position: On the Possibilities of AI-Generated Text Detection | https://proceedings.mlr.press/v235/chakraborty24a.html | Souradip Chakraborty, Amrit Bedi, Sicheng Zhu, Bang An, Dinesh Manocha, Furong Huang | https://proceedings.mlr.press/v235/chakraborty24a.html | ICML 2024 | Our study addresses the challenge of distinguishing human-written text from Large Language Model (LLM) outputs. We provide evidence that this differentiation is consistently feasible, except when human and machine text distributions are indistinguishable across their entire support. Employing information theory, we show that while detecting machine-generated text becomes harder as it nears human quality, it remains possible with adequate text data. We introduce guidelines on the required text data quantity, either through sample size or sequence length, for reliable AI text detection, through derivations of sample complexity bounds. This research paves the way for advanced detection methods. Our comprehensive empirical tests, conducted across various datasets (Xsum, Squad, IMDb, and Kaggle FakeNews) and with several state-of-the-art text generators (GPT-2, GPT-3.5-Turbo, Llama, Llama-2-13B-Chat-HF, Llama-2-70B-Chat-HF), assess the viability of enhanced detection methods against detectors like RoBERTa-Large/Base-Detector and GPTZero, with increasing sample sizes and sequence lengths. Our findings align with OpenAI’s empirical data related to sequence length, marking the first theoretical substantiation for these observations. |
https://proceedings.mlr.press/v235/chakraborty24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chakraborty24b/chakraborty24b.pdf | https://openreview.net/forum?id=8tzjEMF0Vq | MaxMin-RLHF: Alignment with Diverse Human Preferences | https://proceedings.mlr.press/v235/chakraborty24b.html | Souradip Chakraborty, Jiahao Qiu, Hui Yuan, Alec Koppel, Dinesh Manocha, Furong Huang, Amrit Bedi, Mengdi Wang | https://proceedings.mlr.press/v235/chakraborty24b.html | ICML 2024 | Reinforcement Learning from Human Feedback (RLHF) aligns language models to human preferences by employing a singular reward model derived from preference data. However, the single reward model overlooks the rich diversity of human preferences inherent in data collected from multiple users. In this work, we first derive an impossibility result of alignment with single reward RLHF, thereby highlighting its insufficiency in representing diverse human preferences. Next, we propose to learn a mixture of reward models via an expectation-maximization algorithm and solve a MaxMin alignment objective inspired by the Egalitarian principle in social choice theory to better honor diverse human preferences. We present comprehensive experimental results on small-scale (GPT-2) and large-scale language (with Tulu2-7B)) and show the efficacy of the proposed approach in the presence of diversity among human preferences. We remark that our findings in this work are not only limited to language models but also extend to reinforcement learning in general. |
https://proceedings.mlr.press/v235/chan24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chan24a/chan24a.pdf | https://openreview.net/forum?id=eyxVRMrZ4m | Dense Reward for Free in Reinforcement Learning from Human Feedback | https://proceedings.mlr.press/v235/chan24a.html | Alex James Chan, Hao Sun, Samuel Holt, Mihaela Van Der Schaar | https://proceedings.mlr.press/v235/chan24a.html | ICML 2024 | Reinforcement Learning from Human Feedback (RLHF) has been credited as the key advance that has allowed Large Language Models (LLMs) to effectively follow instructions and produce useful assistance. Classically, this involves generating completions from the LLM in response to a query before using a separate reward model to assign a score to the full completion. As an auto-regressive process, the LLM has to take many “actions” (selecting individual tokens) and only receives a single, sparse reward at the end of an episode, a setup that is known to be difficult to optimise in traditional reinforcement learning. In this work we leverage the fact that the reward model contains more information than just its scalar output, in particular, it calculates an attention map over tokens as part of the transformer architecture. We use these attention weights to redistribute the reward along the whole completion, effectively densifying the signal and highlighting the most important tokens, all without incurring extra computational cost or requiring any additional modelling. We demonstrate that, theoretically, this approach is equivalent to potential-based reward shaping, ensuring that the optimal policy remains unchanged. Empirically, we show that it stabilises training, accelerates the rate of learning, and, in practical cases, may lead to better local optima. |
https://proceedings.mlr.press/v235/chan24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chan24b/chan24b.pdf | https://openreview.net/forum?id=Q8uJyOwOsd | Scribble-Supervised Semantic Segmentation with Prototype-based Feature Augmentation | https://proceedings.mlr.press/v235/chan24b.html | Guiyang Chan, Pengcheng Zhang, Hai Dong, Shunhui Ji, Bainian Chen | https://proceedings.mlr.press/v235/chan24b.html | ICML 2024 | Scribble-supervised semantic segmentation presents a cost-effective training method that utilizes annotations generated through scribbling. It is valued in attaining high performance while minimizing annotation costs, which has made it highly regarded among researchers. Scribble supervision propagates information from labeled pixels to the surrounding unlabeled pixels, enabling semantic segmentation for the entire image. However, existing methods often ignore the features of classified pixels during feature propagation. To address these limitations, this paper proposes a prototype-based feature augmentation method that leverages feature prototypes to augment scribble supervision. Experimental results demonstrate that our approach achieves state-of-the-art performance on the PASCAL VOC 2012 dataset in scribble-supervised semantic segmentation tasks. The code is available at https://github.com/TranquilChan/PFA. |
https://proceedings.mlr.press/v235/chang24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chang24a/chang24a.pdf | https://openreview.net/forum?id=fywWm06IGn | Feature Importance Disparities for Data Bias Investigations | https://proceedings.mlr.press/v235/chang24a.html | Peter W Chang, Leor Fishman, Seth Neel | https://proceedings.mlr.press/v235/chang24a.html | ICML 2024 | It is widely held that one cause of downstream bias in classifiers is bias present in the training data. Rectifying such biases may involve context-dependent interventions such as training separate models on subgroups, removing features with bias in the collection process, or even conducting real-world experiments to ascertain sources of bias. Despite the need for such data bias investigations, few automated methods exist to assist practitioners in these efforts. In this paper, we present one such method that given a dataset $X$ consisting of protected and unprotected features, outcomes $y$, and a regressor $h$ that predicts $y$ given $X$, outputs a tuple $(f_j, g)$, with the following property: $g$ corresponds to a subset of the training dataset $(X, y)$, such that the $j^{th}$ feature $f_j$ has much larger (or smaller) influence in the subgroup $g$, than on the dataset overall, which we call feature importance disparity (FID). We show across $4$ datasets and $4$ common feature importance methods of broad interest to the machine learning community that we can efficiently find subgroups with large FID values even over exponentially large subgroup classes and in practice these groups correspond to subgroups with potentially serious bias issues as measured by standard fairness metrics. |
https://proceedings.mlr.press/v235/chang24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chang24b/chang24b.pdf | https://openreview.net/forum?id=KYrAZSbEv6 | Inferring Dynamic Networks from Marginals with Iterative Proportional Fitting | https://proceedings.mlr.press/v235/chang24b.html | Serina Chang, Frederic Koehler, Zhaonan Qu, Jure Leskovec, Johan Ugander | https://proceedings.mlr.press/v235/chang24b.html | ICML 2024 | A common network inference problem, arising from real-world data constraints, is how to infer a dynamic network from its time-aggregated adjacency matrix and time-varying marginals (i.e., row and column sums). Prior approaches to this problem have repurposed the classic iterative proportional fitting (IPF) procedure, also known as Sinkhorn’s algorithm, with promising empirical results. However, the statistical foundation for using IPF has not been well understood: under what settings does IPF provide principled estimation of a dynamic network from its marginals, and how well does it estimate the network? In this work, we establish such a setting, by identifying a generative network model whose maximum likelihood estimates are recovered by IPF. Our model both reveals implicit assumptions on the use of IPF in such settings and enables new analyses, such as structure-dependent error bounds on IPF’s parameter estimates. When IPF fails to converge on sparse network data, we introduce a principled algorithm that guarantees IPF converges under minimal changes to the network structure. Finally, we conduct experiments with synthetic and real-world data, which demonstrate the practical value of our theoretical and algorithmic contributions. |
https://proceedings.mlr.press/v235/chang24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chang24c/chang24c.pdf | https://openreview.net/forum?id=MjGCD8wk1k | LaMAGIC: Language-Model-based Topology Generation for Analog Integrated Circuits | https://proceedings.mlr.press/v235/chang24c.html | Chen-Chia Chang, Yikang Shen, Shaoze Fan, Jing Li, Shun Zhang, Ningyuan Cao, Yiran Chen, Xin Zhang | https://proceedings.mlr.press/v235/chang24c.html | ICML 2024 | In the realm of electronic and electrical engineering, automation of analog circuit is increasingly vital given the complexity and customized requirements of modern applications. However, existing methods only develop search-based algorithms that require many simulation iterations to design a custom circuit topology, which is usually a time-consuming process. To this end, we introduce LaMAGIC, a pioneering language model-based topology generation model that leverages supervised finetuning for automated analog circuit design. LaMAGIC can efficiently generate an optimized circuit design from the custom specification in a single pass. Our approach involves a meticulous development and analysis of various input and output formulations for circuit. These formulations can ensure canonical representations of circuits and align with the autoregressive nature of LMs to effectively addressing the challenges of representing analog circuits as graphs. The experimental results show that LaMAGIC achieves a success rate of up to 96% under a strict tolerance of 0.01. We also examine the scalability and adaptability of LaMAGIC, specifically testing its performance on more complex circuits. Our findings reveal the enhanced effectiveness of our adjacency matrix-based circuit formulation with floating-point input, suggesting its suitability for handling intricate circuit designs. This research not only demonstrates the potential of language models in graph generation, but also builds a foundational framework for future explorations in automated analog circuit design. |
https://proceedings.mlr.press/v235/chang24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chang24d/chang24d.pdf | https://openreview.net/forum?id=jVXJdGQ4eD | MagicPose: Realistic Human Poses and Facial Expressions Retargeting with Identity-aware Diffusion | https://proceedings.mlr.press/v235/chang24d.html | Di Chang, Yichun Shi, Quankai Gao, Hongyi Xu, Jessica Fu, Guoxian Song, Qing Yan, Yizhe Zhu, Xiao Yang, Mohammad Soleymani | https://proceedings.mlr.press/v235/chang24d.html | ICML 2024 | In this work, we propose MagicPose, a diffusion-based model for 2D human pose and facial expression retargeting. Specifically, given a reference image, we aim to generate a person’s new images by controlling the poses and facial expressions while keeping the identity unchanged. To this end, we propose a two-stage training strategy to disentangle human motions and appearance (e.g., facial expressions, skin tone, and dressing), consisting of (1) the pre-training of an appearance-control block and (2) learning appearance-disentangled pose control. Our novel design enables robust appearance control over generated human images, including body, facial attributes, and even background. By leveraging the prior knowledge of image diffusion models, MagicPose generalizes well to unseen human identities and complex poses without the need for additional fine-tuning. Moreover, the proposed model is easy to use and can be considered as a plug-in module/extension to Stable Diffusion. The project website is here. The code is available here. |
https://proceedings.mlr.press/v235/chang24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chang24e/chang24e.pdf | https://openreview.net/forum?id=hTiNFCNxM1 | From Biased Selective Labels to Pseudo-Labels: An Expectation-Maximization Framework for Learning from Biased Decisions | https://proceedings.mlr.press/v235/chang24e.html | Trenton Chang, Jenna Wiens | https://proceedings.mlr.press/v235/chang24e.html | ICML 2024 | Selective labels occur when label observations are subject to a decision-making process; e.g., diagnoses that depend on the administration of laboratory tests. We study a clinically-inspired selective label problem called disparate censorship, where labeling biases vary across subgroups and unlabeled individuals are imputed as “negative” (i.e., no diagnostic test = no illness). Machine learning models naively trained on such labels could amplify labeling bias. Inspired by causal models of selective labels, we propose Disparate Censorship Expectation-Maximization (DCEM), an algorithm for learning in the presence of disparate censorship. We theoretically analyze how DCEM mitigates the effects of disparate censorship on model performance. We validate DCEM on synthetic data, showing that it improves bias mitigation (area between ROC curves) without sacrificing discriminative performance (AUC) compared to baselines. We achieve similar results in a sepsis classification task using clinical data. |
https://proceedings.mlr.press/v235/chanpuriya24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chanpuriya24a/chanpuriya24a.pdf | https://openreview.net/forum?id=0XDO74NlOd | On the Role of Edge Dependency in Graph Generative Models | https://proceedings.mlr.press/v235/chanpuriya24a.html | Sudhanshu Chanpuriya, Cameron N Musco, Konstantinos Sotiropoulos, Charalampos Tsourakakis | https://proceedings.mlr.press/v235/chanpuriya24a.html | ICML 2024 | We investigate the trade-off between the representation power of graph generative models and model overlap, i.e., the degree to which the model generates diverse outputs versus regurgitating its training data. In particular, we delineate a nested hierarchy of graph generative models categorized into three levels of complexity: edge independent, node independent, and arbitrarily dependent models. This hierarchy encapsulates a wide range of prevalent methods. We derive theoretical bounds on the number of triangles and other short-length cycles producible by each level of the hierarchy, finding that more complex dependency structure allows an improved trade-off between representation power and overlap. We provide instances demonstrating the asymptotic optimality of our bounds. Furthermore, we introduce new generative models for each of the three hierarchical levels, leveraging dense subgraph discovery. Our evaluation, conducted on real-world datasets, focuses on assessing the output quality and overlap of our proposed models in comparison to other popular models. Our results indicate that our simple, interpretable models provide competitive baselines to popular generative models. Through this investigation, we offer a structured and robust evaluation scheme, thereby facilitating the development of models capable of generating accurate and edge-diverse graphs. |
https://proceedings.mlr.press/v235/chattopadhyay24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chattopadhyay24a/chattopadhyay24a.pdf | https://openreview.net/forum?id=yTXv8KDD1P | Performance Bounds for Active Binary Testing with Information Maximization | https://proceedings.mlr.press/v235/chattopadhyay24a.html | Aditya Chattopadhyay, Benjamin David Haeffele, Rene Vidal, Donald Geman | https://proceedings.mlr.press/v235/chattopadhyay24a.html | ICML 2024 | In many applications like experimental design, group testing, and medical diagnosis, the state of a random variable $Y$ is revealed by successively observing the outcomes of binary tests about $Y$. New tests are selected adaptively based on the history of outcomes observed so far. If the number of states of $Y$ is finite, the process ends when $Y$ can be predicted with a desired level of confidence or all available tests have been used. Finding the strategy that minimizes the expected number of tests needed to predict $Y$ is virtually impossible in most real applications. Therefore, the commonly used strategy is the greedy heuristic of Information Maximization (InfoMax) that selects tests sequentially in order of information gain. Despite its widespread use, existing guarantees on its performance are often vacuous when compared to its empirical efficiency. In this paper, for the first time to the best of our knowledge, we establish tight non-vacuous bounds on InfoMax’s performance. Our analysis is based on the assumption that at any iteration of the greedy strategy, there is always a binary test available whose conditional probability of being ’true’, given the history, is within $\delta$ units of one-half. This assumption is motivated by practical applications where the available set of tests often satisfies this property for modest values of $\delta$, say, ${0.1 \leq \delta \leq 0.4}$. Specifically, we analyze two distinct scenarios: (i) all tests are functions of $Y$, and (ii) test outcomes are corrupted by a binary symmetric channel. For both cases, our bounds guarantee the near-optimal performance of InfoMax for modest $\delta$ values. It requires only a small multiplicative factor of the entropy of $Y$, in terms of the average number of tests needed to make accurate predictions. |
https://proceedings.mlr.press/v235/che24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/che24a/che24a.pdf | https://openreview.net/forum?id=R6GT1UDcOW | Target Networks and Over-parameterization Stabilize Off-policy Bootstrapping with Function Approximation | https://proceedings.mlr.press/v235/che24a.html | Fengdi Che, Chenjun Xiao, Jincheng Mei, Bo Dai, Ramki Gummadi, Oscar A Ramirez, Christopher K Harris, A. Rupam Mahmood, Dale Schuurmans | https://proceedings.mlr.press/v235/che24a.html | ICML 2024 | We prove that the combination of a target network and over-parameterized linear function approximation establishes a weaker convergence condition for bootstrapped value estimation in certain cases, even with off-policy data. Our condition is naturally satisfied for expected updates over the entire state-action space or learning with a batch of complete trajectories from episodic Markov decision processes. Notably, using only a target network or an over-parameterized model does not provide such a convergence guarantee. Additionally, we extend our results to learning with truncated trajectories, showing that convergence is achievable for all tasks with minor modifications, akin to value truncation for the final states in trajectories. Our primary result focuses on temporal difference estimation for prediction, providing high-probability value estimation error bounds and empirical analysis on Baird’s counterexample and a Four-room task. Furthermore, we explore the control setting, demonstrating that similar convergence conditions apply to Q-learning. |
https://proceedings.mlr.press/v235/chen24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24a/chen24a.pdf | https://openreview.net/forum?id=17ZwoHl65h | PlanDQ: Hierarchical Plan Orchestration via D-Conductor and Q-Performer | https://proceedings.mlr.press/v235/chen24a.html | Chang Chen, Junyeob Baek, Fei Deng, Kenji Kawaguchi, Caglar Gulcehre, Sungjin Ahn | https://proceedings.mlr.press/v235/chen24a.html | ICML 2024 | Despite the recent advancements in offline RL, no unified algorithm could achieve superior performance across a broad range of tasks. Offline value function learning, in particular, struggles with sparse-reward, long-horizon tasks due to the difficulty of solving credit assignment and extrapolation errors that accumulates as the horizon of the task grows. On the other hand, models that can perform well in long-horizon tasks are designed specifically for goal-conditioned tasks, which commonly perform worse than value function learning methods on short-horizon, dense-reward scenarios. To bridge this gap, we propose a hierarchical planner designed for offline RL called PlanDQ. PlanDQ incorporates a diffusion-based planner at the high level, named D-Conductor, which guides the low-level policy through sub-goals. At the low level, we used a Q-learning based approach called the Q-Performer to accomplish these sub-goals. Our experimental results suggest that PlanDQ can achieve superior or competitive performance on D4RL continuous control benchmark tasks as well as AntMaze, Kitchen, and Calvin as long-horizon tasks. |
https://proceedings.mlr.press/v235/chen24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24b/chen24b.pdf | https://openreview.net/forum?id=F3G2udCF3Q | How Interpretable Are Interpretable Graph Neural Networks? | https://proceedings.mlr.press/v235/chen24b.html | Yongqiang Chen, Yatao Bian, Bo Han, James Cheng | https://proceedings.mlr.press/v235/chen24b.html | ICML 2024 | Interpretable graph neural networks (XGNNs ) are widely adopted in various scientific applications involving graph-structured data. Existing XGNNs predominantly adopt the attention-based mechanism to learn edge or node importance for extracting and making predictions with the interpretable subgraph. However, the representational properties and limitations of these methods remain inadequately explored. In this work, we present a theoretical framework that formulates interpretable subgraph learning with the multilinear extension of the subgraph distribution, coined as subgraph multilinear extension (SubMT). Extracting the desired interpretable subgraph requires an accurate approximation of SubMT, yet we find that the existing XGNNs can have a huge gap in fitting SubMT. Consequently, the SubMT approximation failure will lead to the degenerated interpretability of the extracted subgraphs. To mitigate the issue, we design a new XGNN architecture called Graph Multilinear neT (GMT), which is provably more powerful in approximating SubMT. We empirically validate our theoretical findings on a number of graph classification benchmarks. The results demonstrate that GMT outperforms the state-of-the-art up to 10% in terms of both interpretability and generalizability across 12 regular and geometric graph benchmarks. |
https://proceedings.mlr.press/v235/chen24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24c/chen24c.pdf | https://openreview.net/forum?id=5lI9wm4dws | Doubly Robust Causal Effect Estimation under Networked Interference via Targeted Learning | https://proceedings.mlr.press/v235/chen24c.html | Weilin Chen, Ruichu Cai, Zeqin Yang, Jie Qiao, Yuguang Yan, Zijian Li, Zhifeng Hao | https://proceedings.mlr.press/v235/chen24c.html | ICML 2024 | Causal effect estimation under networked interference is an important but challenging problem. Available parametric methods are limited in their model space, while previous semiparametric methods, e.g., leveraging neural networks to fit only one single nuisance function, may still encounter misspecification problems under networked interference without appropriate assumptions on the data generation process. To mitigate bias stemming from misspecification, we propose a novel doubly robust causal effect estimator under networked interference, by adapting the targeted learning technique to the training of neural networks. Specifically, we generalize the targeted learning technique into the networked interference setting and establish the condition under which an estimator achieves double robustness. Based on the condition, we devise an end-to-end causal effect estimator by transforming the identified theoretical condition into a targeted loss. Moreover, we provide a theoretical analysis of our designed estimator, revealing a faster convergence rate compared to a single nuisance model. Extensive experimental results on two real-world networks with semisynthetic data demonstrate the effectiveness of our proposed estimators. |
https://proceedings.mlr.press/v235/chen24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24d/chen24d.pdf | https://openreview.net/forum?id=J6prHJsIlf | Feature Attribution with Necessity and Sufficiency via Dual-stage Perturbation Test for Causal Explanation | https://proceedings.mlr.press/v235/chen24d.html | Xuexin Chen, Ruichu Cai, Zhengting Huang, Yuxuan Zhu, Julien Horwood, Zhifeng Hao, Zijian Li, José Miguel Hernández-Lobato | https://proceedings.mlr.press/v235/chen24d.html | ICML 2024 | We investigate the problem of explainability for machine learning models, focusing on Feature Attribution Methods (FAMs) that evaluate feature importance through perturbation tests. Despite their utility, FAMs struggle to distinguish the contributions of different features, when their prediction changes are similar after perturbation. To enhance FAMs’ discriminative power, we introduce Feature Attribution with Necessity and Sufficiency (FANS), which find a neighborhood of the input such that perturbing samples within this neighborhood have a high Probability of being Necessity and Sufficiency (PNS) cause for the change in predictions, and use this PNS as the importance of the feature. Specifically, FANS compute this PNS via a heuristic strategy for estimating the neighborhood and a perturbation test involving two stages (factual and interventional) for counterfactual reasoning. To generate counterfactual samples, we use a resampling-based approach on the observed samples to approximate the required conditional distribution. We demonstrate that FANS outperforms existing attribution methods on six benchmarks. Please refer to the source code via https://github.com/DMIRLAB-Group/FANS. |
https://proceedings.mlr.press/v235/chen24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24e/chen24e.pdf | https://openreview.net/forum?id=rADFNrIss3 | InstructZero: Efficient Instruction Optimization for Black-Box Large Language Models | https://proceedings.mlr.press/v235/chen24e.html | Lichang Chen, Jiuhai Chen, Tom Goldstein, Heng Huang, Tianyi Zhou | https://proceedings.mlr.press/v235/chen24e.html | ICML 2024 | Large language models (LLMs) are instruction followers but the performance varies under different instructions. It is challenging to create the best instruction, especially for black-box LLMs on which backpropagation is forbidden. Instead of directly optimizing the discrete instruction, we optimize a low-dimensional soft prompt applied to an open-source LLM to generate the instruction for the black-box LLM. In each optimization step of the proposed method InstructZero, a soft prompt is converted into an instruction by the open-source LLM, which is then submitted to the black-box LLM for zero-shot evaluation, whose result is sent to Bayesian optimization to produce new soft prompts improving the zero-shot performance. We evaluate InstructZero on different combinations of open-source LLMs and APIs including Vicuna and ChatGPT. InstructZero outperforms SOTA auto-instruction methods across a variety of downstream tasks. |
https://proceedings.mlr.press/v235/chen24f.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24f/chen24f.pdf | https://openreview.net/forum?id=61RlaY9EIn | MaSS: Multi-attribute Selective Suppression for Utility-preserving Data Transformation from an Information-theoretic Perspective | https://proceedings.mlr.press/v235/chen24f.html | Yizhuo Chen, Chun-Fu Chen, Hsiang Hsu, Shaohan Hu, Marco Pistoia, Tarek F. Abdelzaher | https://proceedings.mlr.press/v235/chen24f.html | ICML 2024 | The growing richness of large-scale datasets has been crucial in driving the rapid advancement and wide adoption of machine learning technologies. The massive collection and usage of data, however, pose an increasing risk for people’s private and sensitive information due to either inadvertent mishandling or malicious exploitation. Besides legislative solutions, many technical approaches have been proposed towards data privacy protection. However, they bear various limitations such as leading to degraded data availability and utility, or relying on heuristics and lacking solid theoretical bases. To overcome these limitations, we propose a formal information-theoretic definition for this utility-preserving privacy protection problem, and design a data-driven learnable data transformation framework that is capable of selectively suppressing sensitive attributes from target datasets while preserving the other useful attributes, regardless of whether or not they are known in advance or explicitly annotated for preservation. We provide rigorous theoretical analyses on the operational bounds for our framework, and carry out comprehensive experimental evaluations using datasets of a variety of modalities, including facial images, voice audio clips, and human activity motion sensor signals. Results demonstrate the effectiveness and generalizability of our method under various configurations on a multitude of tasks. Our source code is available at this URL. |
https://proceedings.mlr.press/v235/chen24g.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24g/chen24g.pdf | https://openreview.net/forum?id=g9mYBdooPA | Policy-conditioned Environment Models are More Generalizable | https://proceedings.mlr.press/v235/chen24g.html | Ruifeng Chen, Xiong-Hui Chen, Yihao Sun, Siyuan Xiao, Minhui Li, Yang Yu | https://proceedings.mlr.press/v235/chen24g.html | ICML 2024 | In reinforcement learning, it is crucial to have an accurate environment dynamics model to evaluate different policies’ value in downstream tasks like offline policy optimization and policy evaluation. However, the learned model is known to be inaccurate in predictions when evaluating target policies different from data-collection policies. In this work, we found that utilizing policy representation for model learning, called policy-conditioned model (PCM) learning, is useful to mitigate the problem, especially when the offline dataset is collected from diversified behavior policies. The reason beyond that is in this case, PCM becomes a meta-dynamics model that is trained to be aware of and focus on the evaluation policies that on-the-fly adjust the model to be suitable to the evaluation policies’ state-action distribution, thus improving the prediction accuracy. Based on that intuition, we propose an easy-to-implement yet effective algorithm of PCM for accurate model learning. We also give a theoretical analysis and experimental evidence to demonstrate the feasibility of reducing value gaps by adapting the dynamics model under different policies. Experiment results show that PCM outperforms the existing SOTA off-policy evaluation methods in the DOPE benchmark by a large margin, and derives significantly better policies in offline policy selection and model predictive control compared with the standard model learning method. |
https://proceedings.mlr.press/v235/chen24h.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24h/chen24h.pdf | https://openreview.net/forum?id=dbFEFHAD79 | MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmark | https://proceedings.mlr.press/v235/chen24h.html | Dongping Chen, Ruoxi Chen, Shilin Zhang, Yaochen Wang, Yinuo Liu, Huichi Zhou, Qihui Zhang, Yao Wan, Pan Zhou, Lichao Sun | https://proceedings.mlr.press/v235/chen24h.html | ICML 2024 | Multimodal Large Language Models (MLLMs) have gained significant attention recently, showing remarkable potential in artificial general intelligence. However, assessing the utility of MLLMs presents considerable challenges, primarily due to the absence multimodal benchmarks that align with human preferences. Drawing inspiration from the concept of LLM-as-a-Judge within LLMs, this paper introduces a novel benchmark, termed MLLM-as-a-Judge, to assess the ability of MLLMs in assisting judges across diverse modalities, encompassing three distinct tasks: Scoring Evaluation, Pair Comparison, and Batch Ranking. Our study reveals that, while MLLMs demonstrate remarkable human-like discernment in Pair Comparisons, there is a significant divergence from human preferences in Scoring Evaluation and Batch Ranking tasks. Furthermore, a closer examination reveals persistent challenges in the evaluative capacities of LLMs, including diverse biases, hallucinatory responses, and inconsistencies in judgment, even in advanced models such as GPT-4V. These findings emphasize the pressing need for enhancements and further research efforts to be undertaken before regarding MLLMs as fully reliable evaluators. In light of this, we advocate for additional efforts dedicated to supporting the continuous development within the domain of MLLM functioning as judges. The code and dataset are publicly available at our project homepage: https://mllm-judge.github.io/. |
https://proceedings.mlr.press/v235/chen24i.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24i/chen24i.pdf | https://openreview.net/forum?id=4zAHgkiCQg | Premise Order Matters in Reasoning with Large Language Models | https://proceedings.mlr.press/v235/chen24i.html | Xinyun Chen, Ryan Andrew Chi, Xuezhi Wang, Denny Zhou | https://proceedings.mlr.press/v235/chen24i.html | ICML 2024 | Large language models (LLMs) have accomplished remarkable reasoning performance in various domains. However, in the domain of reasoning tasks, we discover a frailty: LLMs are surprisingly brittle to the ordering of the premises, despite the fact that such ordering does not alter the underlying task. In particular, we observe that LLMs achieve the best performance when the premise order aligns with the context required in intermediate reasoning steps. For example, in deductive reasoning tasks, presenting the premises in the same order as the ground truth proof in the prompt (as opposed to random ordering) drastically increases the model’s accuracy. We first examine the effect of premise ordering on deductive reasoning on a variety of LLMs, and our evaluation shows that even if the model performance is decent on the optimal order, permuting the premise order can cause a performance drop of over 30%. In addition, we release the benchmark R-GSM, based on GSM8K, to examine the ordering effect for mathematical problem-solving, and we again observe a significant drop in accuracy, relative to the original GSM8K benchmark. |
https://proceedings.mlr.press/v235/chen24j.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24j/chen24j.pdf | https://openreview.net/forum?id=O4cHTxW9BS | Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models | https://proceedings.mlr.press/v235/chen24j.html | Zixiang Chen, Yihe Deng, Huizhuo Yuan, Kaixuan Ji, Quanquan Gu | https://proceedings.mlr.press/v235/chen24j.html | ICML 2024 | Harnessing the power of human-annotated data through Supervised Fine-Tuning (SFT) is pivotal for advancing Large Language Models (LLMs). In this paper, we delve into the prospect of growing a strong LLM out of a weak one without the need for acquiring additional human-annotated data. We propose a new fine-tuning method called Self-Play fIne-tuNing (SPIN), which starts from a supervised fine-tuned model. At the heart of SPIN lies a self-play mechanism, where the LLM refines its capability by playing against instances of itself. More specifically, the LLM generates its own training data from its previous iterations, refining its policy by discerning these self-generated responses from those obtained from human-annotated data. Our method progressively elevates the LLM from a nascent model to a formidable one, unlocking the full potential of human-annotated demonstration data for SFT. Theoretically, we prove that the global optimum to the training objective function of our method is achieved only when the LLM policy aligns with the target data distribution. Empirically, we evaluate our method on several benchmark datasets including the HuggingFace Open LLM Leaderboard, MT-Bench, and datasets from Big-Bench. Our results show that SPIN can significantly improve the LLM’s performance across a variety of benchmarks and even outperform models trained through direct preference optimization (DPO) supplemented with extra GPT-4 preference data. This sheds light on the promise of self-play, enabling the achievement of human-level performance in LLMs without the need for expert opponents. |
https://proceedings.mlr.press/v235/chen24k.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24k/chen24k.pdf | https://openreview.net/forum?id=xaSpuvNYwS | Robust Classification via a Single Diffusion Model | https://proceedings.mlr.press/v235/chen24k.html | Huanran Chen, Yinpeng Dong, Zhengyi Wang, Xiao Yang, Chengqi Duan, Hang Su, Jun Zhu | https://proceedings.mlr.press/v235/chen24k.html | ICML 2024 | Diffusion models have been applied to improve adversarial robustness of image classifiers by purifying the adversarial noises or generating realistic data for adversarial training. However, diffusion-based purification can be evaded by stronger adaptive attacks while adversarial training does not perform well under unseen threats, exhibiting inevitable limitations of these methods. To better harness the expressive power of diffusion models, this paper proposes Robust Diffusion Classifier (RDC), a generative classifier that is constructed from a pre-trained diffusion model to be adversarially robust. RDC first maximizes the data likelihood of a given input and then predicts the class probabilities of the optimized input using the conditional likelihood estimated by the diffusion model through Bayes’ theorem. To further reduce the computational cost, we propose a new diffusion backbone called multi-head diffusion and develop efficient sampling strategies. As RDC does not require training on particular adversarial attacks, we demonstrate that it is more generalizable to defend against multiple unseen threats. In particular, RDC achieves $75.67%$ robust accuracy against various $\ell_\infty$ norm-bounded adaptive attacks with $\epsilon_\infty=8/255$ on CIFAR-10, surpassing the previous state-of-the-art adversarial training models by $+4.77%$. The results highlight the potential of generative classifiers by employing pre-trained diffusion models for adversarial robustness compared with the commonly studied discriminative classifiers. |
https://proceedings.mlr.press/v235/chen24l.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24l/chen24l.pdf | https://openreview.net/forum?id=puSMYmHmJW | Relational Learning in Pre-Trained Models: A Theory from Hypergraph Recovery Perspective | https://proceedings.mlr.press/v235/chen24l.html | Yang Chen, Cong Fang, Zhouchen Lin, Bing Liu | https://proceedings.mlr.press/v235/chen24l.html | ICML 2024 | Foundation Models (FMs) have demonstrated remarkable insights into the relational dynamics of the world, leading to the crucial question: how do these models acquire an understanding of world hybrid relations? Traditional statistical learning, particularly for prediction problems, may overlook the rich and inherently structured information from the data, especially regarding the relationships between objects. We introduce a mathematical model that formalizes relational learning as hypergraph recovery to study pre-training of FMs. In our framework, the world is represented as a hypergraph, with data abstracted as random samples from hyperedges. We theoretically examine the feasibility of a Pre-Trained Model (PTM) to recover this hypergraph and analyze the data efficiency in a minimax near-optimal style. By integrating rich graph theories into the realm of PTMs, our mathematical framework offers powerful tools for an in-depth understanding of pre-training from a unique perspective and can be used under various scenarios. As an example, we extend the framework to entity alignment in multimodal learning. |
https://proceedings.mlr.press/v235/chen24m.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24m/chen24m.pdf | https://openreview.net/forum?id=JVhUR8q27o | Towards AutoAI: Optimizing a Machine Learning System with Black-box and Differentiable Components | https://proceedings.mlr.press/v235/chen24m.html | Zhiliang Chen, Chuan-Sheng Foo, Bryan Kian Hsiang Low | https://proceedings.mlr.press/v235/chen24m.html | ICML 2024 | Machine learning (ML) models in the real world typically do not exist in isolation. They are usually part of a complex system (e.g., healthcare systems, self-driving cars) containing multiple ML and black-box components. The problem of optimizing such systems, which we refer to as automated AI (AutoAI), requires us to jointly train all ML components together and presents a significant challenge because the number of system parameters is extremely high and the system has no analytical form. To circumvent this, we introduce a novel algorithm called A-BAD-BO which uses each ML component’s local loss as an auxiliary indicator for system performance. A-BAD-BO uses Bayesian optimization (BO) to optimize the local loss configuration of a system in a smaller dimensional space and exploits the differentiable structure of ML components to recover optimal system parameters from the optimized configuration. We show A-BAD-BO converges to optimal system parameters by showing that it is asymptotically no regret. We use A-BAD-BO to optimize several synthetic and real-world complex systems, including a prompt engineering pipeline for large language models containing millions of system parameters. Our results demonstrate that A-BAD-BO yields better system optimality than gradient-driven baselines and is more sample-efficient than pure BO algorithms. |
https://proceedings.mlr.press/v235/chen24n.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24n/chen24n.pdf | https://openreview.net/forum?id=UQYXZdca92 | Probabilistic Forecasting with Stochastic Interpolants and Föllmer Processes | https://proceedings.mlr.press/v235/chen24n.html | Yifan Chen, Mark Goldstein, Mengjian Hua, Michael Samuel Albergo, Nicholas Matthew Boffi, Eric Vanden-Eijnden | https://proceedings.mlr.press/v235/chen24n.html | ICML 2024 | We propose a framework for probabilistic forecasting of dynamical systems based on generative modeling. Given observations of the system state over time, we formulate the forecasting problem as sampling from the conditional distribution of the future system state given its current state. To this end, we leverage the framework of stochastic interpolants, which facilitates the construction of a generative model between an arbitrary base distribution and the target. We design a fictitious, non-physical stochastic dynamics that takes as initial condition the current system state and produces as output a sample from the target conditional distribution in finite time and without bias. This process therefore maps a point mass centered at the current state onto a probabilistic ensemble of forecasts. We prove that the drift coefficient entering the stochastic differential equation (SDE) achieving this task is non-singular, and that it can be learned efficiently by square loss regression over the time-series data. We show that the drift and the diffusion coefficients of this SDE can be adjusted after training, and that a specific choice that minimizes the impact of the estimation error gives a Föllmer process. We highlight the utility of our approach on several complex, high-dimensional forecasting problems, including stochastically forced Navier-Stokes and video prediction on the KTH and CLEVRER datasets. The code is available at https://github.com/interpolants/forecasting. |
https://proceedings.mlr.press/v235/chen24o.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24o/chen24o.pdf | https://openreview.net/forum?id=iLSgF7jMtI | CogDPM: Diffusion Probabilistic Models via Cognitive Predictive Coding | https://proceedings.mlr.press/v235/chen24o.html | Kaiyuan Chen, Xingzhuo Guo, Yu Zhang, Jianmin Wang, Mingsheng Long | https://proceedings.mlr.press/v235/chen24o.html | ICML 2024 | Predictive Coding (PC) is a theoretical framework in cognitive science suggesting that the human brain processes cognition through spatiotemporal prediction of visual world. Existing studies have developed spatiotemporal prediction neural networks based on the PC theroy, emulating its two core mechanisms: Correcting predictions from residuals and Hierarchical learning. However, these models do not show the enhancement of prediction skills on real-world forecasting tasks, and ignore the Precision Weighting mechanism of PC theory. Precision weight posits that the brain allocates more attention to signals with lower Precision, contributing to the the cognitive ability of human brains. This work introduces the Cognitive Diffusion Probabilistic Models (CogDPM) which demonstrates the connection between diffusion probabilistic models and PC theory. CogDPM features a precision estimation method based on the hierarchical sampling capabilities of diffusion models, and allocate the guidance with precision weights estimated by the inherent property of diffusion models. We experimentally show that the precision weights is an estimator of model’s predictability on the rigid body and fluid motion dataset. We also apply CogDPM to real-world prediction tasks using the U.K. precipitation and ERA surface wind datasets. Our results demonstrate that CogDPM outperforms both existing domain-specific operational models and general deep prediction models in providing more proficient forecasting. |
https://proceedings.mlr.press/v235/chen24p.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24p/chen24p.pdf | https://openreview.net/forum?id=qIiPM5CbRY | On Interpolating Experts and Multi-Armed Bandits | https://proceedings.mlr.press/v235/chen24p.html | Houshuang Chen, Yuchen He, Chihao Zhang | https://proceedings.mlr.press/v235/chen24p.html | ICML 2024 | Learning with expert advice and multi-armed bandit are two classic online decision problems which differ on how the information is observed in each round of the game. We study a family of problems interpolating the two. For a vector $\mathbf{m}=(m_1,…,m_K)\in \mathbb N^K$, an instance of $\mathbf m$-MAB indicates that the arms are partitioned into $K$ groups and the $i$-th group contains $m_i$ arms. Once an arm is pulled, the losses of all arms in the same group are observed. We prove tight minimax regret bounds for $\mathbf m$-MAB and design an optimal PAC algorithm for its pure exploration version, $\mathbf m$-BAI, where the goal is to identify the arm with minimum loss with as few rounds as possible. We show that the minimax regret of $\mathbf m$-MAB is $\Theta\left(\sqrt{T\sum_{k=1}^K\log (m_k+1)}\right)$ and the minimum number of pulls for an $(\varepsilon,0.05)$-PAC algorithm of $\mathbf m$-BAI is $\Theta\left(\frac{1}{\varepsilon^2}\cdot \sum_{k=1}^K\log (m_k+1)\right)$. Both our upper bounds and lower bounds for $\mathbf m$-MAB can be extended to a more general setting, namely the bandit with graph feedback, in terms of the clique cover and related graph parameters. As consequences, we obtained tight minimax regret bounds for several families of feedback graphs. |
https://proceedings.mlr.press/v235/chen24q.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24q/chen24q.pdf | https://openreview.net/forum?id=IoUOhnCmlX | Bagged Deep Image Prior for Recovering Images in the Presence of Speckle Noise | https://proceedings.mlr.press/v235/chen24q.html | Xi Chen, Zhewen Hou, Christopher Metzler, Arian Maleki, Shirin Jalali | https://proceedings.mlr.press/v235/chen24q.html | ICML 2024 | We investigate both the theoretical and algorithmic aspects of likelihood-based methods for recovering a complex-valued signal from multiple sets of measurements, referred to as looks, affected by speckle (multiplicative) noise. Our theoretical contributions include establishing the first existing theoretical upper bound on the Mean Squared Error (MSE) of the maximum likelihood estimator under the deep image prior hypothesis. Our theoretical results capture the dependence of MSE upon the number of parameters in the deep image prior, the number of looks, the signal dimension, and the number of measurements per look. On the algorithmic side, we introduce the concept of bagged Deep Image Priors (Bagged-DIP) and integrate them with projected gradient descent. Furthermore, we show how employing Newton-Schulz algorithm for calculating matrix inverses within the iterations of PGD reduces the computational complexity of the algorithm. We will show that this method achieves the state-of-the-art performance. |
https://proceedings.mlr.press/v235/chen24r.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24r/chen24r.pdf | https://openreview.net/forum?id=LVF4P1NNwO | Exact Conversion of In-Context Learning to Model Weights in Linearized-Attention Transformers | https://proceedings.mlr.press/v235/chen24r.html | Brian K Chen, Tianyang Hu, Hui Jin, Hwee Kuan Lee, Kenji Kawaguchi | https://proceedings.mlr.press/v235/chen24r.html | ICML 2024 | In-Context Learning (ICL) has been a powerful emergent property of large language models that has attracted increasing attention in recent years. In contrast to regular gradient-based learning, ICL is highly interpretable and does not require parameter updates. In this paper, we show that, for linearized transformer networks, ICL can be made explicit and permanent through the inclusion of bias terms. We mathematically demonstrate the equivalence between a model with ICL demonstration prompts and the same model with the additional bias terms. Our algorithm (ICLCA) allows for exact conversion in an inexpensive manner. Existing methods are not exact and require expensive parameter updates. We demonstrate the efficacy of our approach through experiments that show the exact incorporation of ICL tokens into a linear transformer. We further suggest how our method can be adapted to achieve cheap approximate conversion of ICL tokens, even in regular transformer networks that are not linearized. Our experiments on GPT-2 show that, even though the conversion is only approximate, the model still gains valuable context from the included bias terms. |
https://proceedings.mlr.press/v235/chen24s.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24s/chen24s.pdf | https://openreview.net/forum?id=J16WEPdqhJ | Accelerated Policy Gradient for s-rectangular Robust MDPs with Large State Spaces | https://proceedings.mlr.press/v235/chen24s.html | Ziyi Chen, Heng Huang | https://proceedings.mlr.press/v235/chen24s.html | ICML 2024 | Robust Markov decision process (robust MDP) is an important machine learning framework to make a reliable policy that is robust to environmental perturbation. Despite empirical success and popularity of policy gradient methods, existing policy gradient methods require at least iteration complexity $\mathcal{O}(\epsilon^{-4})$ to converge to the global optimal solution of s-rectangular robust MDPs with $\epsilon$-accuracy and are limited to deterministic setting with access to exact gradients and small state space that are impractical in many applications. In this work, we propose an accelerated policy gradient algorithm with iteration complexity $\mathcal{O}(\epsilon^{-3}\ln\epsilon^{-1})$ in the deterministic setting using entropy regularization. Furthermore, we extend this algorithm to stochastic setting with access to only stochastic gradients and large state space which achieves the sample complexity $\mathcal{O}(\epsilon^{-7}\ln\epsilon^{-1})$. In the meantime, our algorithms are also the first scalable policy gradient methods to entropy-regularized robust MDPs, which provide an important but underexplored machine learning framework. |
https://proceedings.mlr.press/v235/chen24t.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24t/chen24t.pdf | https://openreview.net/forum?id=aeXRBnLoPP | Accelerated Policy Gradient: On the Convergence Rates of the Nesterov Momentum for Reinforcement Learning | https://proceedings.mlr.press/v235/chen24t.html | Yen-Ju Chen, Nai-Chieh Huang, Ching-Pei Lee, Ping-Chun Hsieh | https://proceedings.mlr.press/v235/chen24t.html | ICML 2024 | Various acceleration approaches for Policy Gradient (PG) have been analyzed within the realm of Reinforcement Learning (RL). However, the theoretical understanding of the widely used momentum-based acceleration method on PG remains largely open. In response to this gap, we adapt the celebrated Nesterov’s accelerated gradient (NAG) method to policy optimization in RL, termed Accelerated Policy Gradient (APG). To demonstrate the potential of APG in achieving fast convergence, we formally prove that with the true gradient and under the softmax policy parametrization, APG converges to an optimal policy at rates: (i) $\tilde{O}(1/t^2)$ with nearly constant step sizes; (ii) $O(e^{-ct})$ with time-varying step sizes. To the best of our knowledge, this is the first characterization of the convergence rates of NAG in the context of RL. Notably, our analysis relies on one interesting finding: Regardless of the parameter initialization, APG ends up entering a locally nearly-concave regime, where APG can significantly benefit from the momentum, within finite iterations. Through numerical validation and experiments on the Atari 2600 benchmarks, we confirm that APG exhibits a $\tilde{O}(1/t^2)$ rate with nearly constant step sizes and a linear convergence rate with time-varying step sizes, significantly improving convergence over the standard PG. |
https://proceedings.mlr.press/v235/chen24u.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24u/chen24u.pdf | https://openreview.net/forum?id=d2vONO90Rw | From Yes-Men to Truth-Tellers: Addressing Sycophancy in Large Language Models with Pinpoint Tuning | https://proceedings.mlr.press/v235/chen24u.html | Wei Chen, Zhen Huang, Liang Xie, Binbin Lin, Houqiang Li, Le Lu, Xinmei Tian, Deng Cai, Yonggang Zhang, Wenxiao Wang, Xu Shen, Jieping Ye | https://proceedings.mlr.press/v235/chen24u.html | ICML 2024 | Large Language Models (LLMs) tend to prioritize adherence to user prompts over providing veracious responses, leading to the sycophancy issue. When challenged by users, LLMs tend to admit mistakes and provide inaccurate responses even if they initially provided the correct answer. Recent works propose to employ supervised fine-tuning (SFT) to mitigate the sycophancy issue, while it typically leads to the degeneration of LLMs’ general capability. To address the challenge, we propose a novel supervised pinpoint tuning (SPT), where the region-of-interest modules are tuned for a given objective. Specifically, SPT first reveals and verifies a small percentage ($<$5%) of the basic modules, which significantly affect a particular behavior of LLMs. i.e., sycophancy. Subsequently, SPT merely fine-tunes these identified modules while freezing the rest. To verify the effectiveness of the proposed SPT, we conduct comprehensive experiments, demonstrating that SPT significantly mitigates the sycophancy issue of LLMs (even better than SFT). Moreover, SPT introduces limited or even no side effects on the general capability of LLMs. Our results shed light on how to precisely, effectively, and efficiently explain and improve the targeted ability of LLMs. |
https://proceedings.mlr.press/v235/chen24v.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24v/chen24v.pdf | https://openreview.net/forum?id=x1G7ieRgRd | Improved Communication-Privacy Trade-offs in $L_2$ Mean Estimation under Streaming Differential Privacy | https://proceedings.mlr.press/v235/chen24v.html | Wei-Ning Chen, Berivan Isik, Peter Kairouz, Albert No, Sewoong Oh, Zheng Xu | https://proceedings.mlr.press/v235/chen24v.html | ICML 2024 | We study $L_2$ mean estimation under central differential privacy and communication constraints, and address two key challenges: firstly, existing mean estimation schemes that simultaneously handle both constraints are usually optimized for $L_\infty$ geometry and rely on random rotation or Kashin’s representation to adapt to $L_2$ geometry, resulting in suboptimal leading constants in mean square errors (MSEs); secondly, schemes achieving order-optimal communication-privacy trade-offs do not extend seamlessly to streaming differential privacy (DP) settings (e.g., tree aggregation or matrix factorization), rendering them incompatible with DP-FTRL type optimizers. In this work, we tackle these issues by introducing a novel privacy accounting method for the sparsified Gaussian mechanism that incorporates the randomness inherent in sparsification into the DP noise. Unlike previous approaches, our accounting algorithm directly operates in $L_2$ geometry, yielding MSEs that fast converge to those of the uncompressed Gaussian mechanism. Additionally, we extend the sparsification scheme to the matrix factorization framework under streaming DP and provide a precise accountant tailored for DP-FTRL type optimizers. Empirically, our method demonstrates at least a 100x improvement of compression for DP-SGD across various FL tasks. |
https://proceedings.mlr.press/v235/chen24w.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24w/chen24w.pdf | https://openreview.net/forum?id=dqpg8jdA2w | Offline Transition Modeling via Contrastive Energy Learning | https://proceedings.mlr.press/v235/chen24w.html | Ruifeng Chen, Chengxing Jia, Zefang Huang, Tian-Shuo Liu, Xu-Hui Liu, Yang Yu | https://proceedings.mlr.press/v235/chen24w.html | ICML 2024 | Learning a high-quality transition model is of great importance for sequential decision-making tasks, especially in offline settings. Nevertheless, the complex behaviors of transition dynamics in real-world environments pose challenges for the standard forward models because of their inductive bias towards smooth regressors, conflicting with the inherent nature of transitions such as discontinuity or large curvature. In this work, we propose to model the transition probability implicitly through a scalar-value energy function, which enables not only flexible distribution prediction but also capturing complex transition behaviors. The Energy-based Transition Models (ETM) are shown to accurately fit the discontinuous transition functions and better generalize to out-of-distribution transition data. Furthermore, we demonstrate that energy-based transition models improve the evaluation accuracy and significantly outperform other off-policy evaluation methods in DOPE benchmark. Finally, we show that energy-based transition models also benefit reinforcement learning and outperform prior offline RL algorithms in D4RL Gym-Mujoco tasks. |
https://proceedings.mlr.press/v235/chen24x.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24x/chen24x.pdf | https://openreview.net/forum?id=a2uFstsHPb | Efficient Pareto Manifold Learning with Low-Rank Structure | https://proceedings.mlr.press/v235/chen24x.html | Weiyu Chen, James Kwok | https://proceedings.mlr.press/v235/chen24x.html | ICML 2024 | Multi-task learning, which optimizes performance across multiple tasks, is inherently a multi-objective optimization problem. Various algorithms are developed to provide discrete trade-off solutions on the Pareto front. Recently, continuous Pareto front approximations using a linear combination of base networks have emerged as a compelling strategy. However, it suffers from scalability issues when the number of tasks is large. To address this issue, we propose a novel approach that integrates a main network with several low-rank matrices to efficiently learn the Pareto manifold. It significantly reduces the number of parameters and facilitates the extraction of shared features. We also introduce orthogonal regularization to further bolster performance. Extensive experimental results demonstrate that the proposed approach outperforms state-of-the-art baselines, especially on datasets with a large number of tasks. |
https://proceedings.mlr.press/v235/chen24y.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24y/chen24y.pdf | https://openreview.net/forum?id=aoAPOOtN9E | Toward Adaptive Reasoning in Large Language Models with Thought Rollback | https://proceedings.mlr.press/v235/chen24y.html | Sijia Chen, Baochun Li | https://proceedings.mlr.press/v235/chen24y.html | ICML 2024 | Large language models (LLMs) have been routinely used to solve various tasks using step-by-step reasoning. However, the structure of intermediate reasoning steps, or thoughts, is rigid and unidirectional, such as chains, trees, or acyclic-directed graphs. Consequently, the resulting inflexible and forward-only reasoning may not address challenging tasks and fail when the LLM frequently gives false responses, i.e., hallucinations. This paper proposes a new reasoning framework, called Thought Rollback (TR), allowing LLMs to adaptively build thought structure while maintaining effective reasoning toward problem-solving under hallucinations. The core mechanism of TR is rolling back thoughts, which allows LLMs to perform error analysis on thoughts, and thus roll back to any previously mistaken thought for revision. Subsequently, by including such trial-and-error in the prompt to guide the LLM, each rollback leads to one more reliable reasoning path. Therefore, starting with a simple prompt without human annotations, LLM with TR adaptively and gradually explores thoughts for a correct solution. Comprehensive experiments on mathematical problems and multi-task reasoning demonstrate the state-of-the-art performance of TR in terms of problem-solving rate and interaction cost. For instance, the solving rate of GPT-4 with TR outperforms the current best by $9%$ on the MATH dataset. The source code is available under the folder examples/ThoughtRollback of https://github.com/iQua/llmpebase. |
https://proceedings.mlr.press/v235/chen24z.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24z/chen24z.pdf | https://openreview.net/forum?id=JU3xHh1vWw | Identifiability Matters: Revealing the Hidden Recoverable Condition in Unbiased Learning to Rank | https://proceedings.mlr.press/v235/chen24z.html | Mouxiang Chen, Chenghao Liu, Zemin Liu, Zhuo Li, Jianling Sun | https://proceedings.mlr.press/v235/chen24z.html | ICML 2024 | Unbiased Learning to Rank (ULTR) aims to train unbiased ranking models from biased click logs, by explicitly modeling a generation process for user behavior and fitting click data based on examination hypothesis. Previous research found empirically that the true latent relevance is mostly recoverable through click fitting. However, we demonstrate that this is not always achievable, resulting in a significant reduction in ranking performance. This research investigates the conditions under which relevance can be recovered from click data in the first principle. We initially characterize a ranking model as identifiable if it can recover the true relevance up to a scaling transformation, a criterion sufficient for the pairwise ranking objective. Subsequently, we investigate an equivalent condition for identifiability, articulated as a graph connectivity test problem: the recovery of relevance is feasible if and only if the identifiability graph (IG), derived from the underlying structure of the dataset, is connected. The presence of a disconnected IG may lead to degenerate cases and suboptimal ranking performance. To tackle this challenge, we introduce two methods, namely node intervention and node merging, designed to modify the dataset and restore the connectivity of the IG. Empirical results derived from a simulated dataset and two real-world LTR benchmark datasets not only validate our proposed theory, but also demonstrate the effectiveness of our methods in alleviating data bias when the relevance model is unidentifiable. |
https://proceedings.mlr.press/v235/chen24aa.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24aa/chen24aa.pdf | https://openreview.net/forum?id=bBzlapzeR1 | High-Dimensional Kernel Methods under Covariate Shift: Data-Dependent Implicit Regularization | https://proceedings.mlr.press/v235/chen24aa.html | Yihang Chen, Fanghui Liu, Taiji Suzuki, Volkan Cevher | https://proceedings.mlr.press/v235/chen24aa.html | ICML 2024 | This paper studies kernel ridge regression in high dimensions under covariate shifts and analyzes the role of importance re-weighting. We first derive the asymptotic expansion of high dimensional kernels under covariate shifts. By a bias-variance decomposition, we theoretically demonstrate that the re-weighting strategy allows for decreasing the variance. For bias, we analyze the regularization of the arbitrary or well-chosen scale, showing that the bias can behave very differently under different regularization scales. In our analysis, the bias and variance can be characterized by the spectral decay of a data-dependent regularized kernel: the original kernel matrix associated with an additional re-weighting matrix, and thus the re-weighting strategy can be regarded as a data-dependent regularization for better understanding. Besides, our analysis provides asymptotic expansion of kernel functions/vectors under covariate shift, which has its own interest. |
https://proceedings.mlr.press/v235/chen24ab.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ab/chen24ab.pdf | https://openreview.net/forum?id=0uUHfhXdnH | DiJiang: Efficient Large Language Models through Compact Kernelization | https://proceedings.mlr.press/v235/chen24ab.html | Hanting Chen, Liu Zhicheng, Xutao Wang, Yuchuan Tian, Yunhe Wang | https://proceedings.mlr.press/v235/chen24ab.html | ICML 2024 | In an effort to reduce the computational load of Transformers, research on linear attention has gained significant momentum. However, the improvement strategies for attention mechanisms typically necessitate extensive retraining, which is impractical for large language models with a vast array of parameters. In this paper, we present DiJiang, a novel Frequency Domain Kernelization approach that enables the transformation of a pre-trained vanilla Transformer into a linear complexity model with little training costs. By employing a weighted Quasi-Monte Carlo method for sampling, the proposed approach theoretically offers superior approximation efficiency. To further reduce the training computational complexity, our kernelization is based on Discrete Cosine Transform (DCT) operations. Extensive experiments demonstrate that the proposed method achieves comparable performance to the original Transformer, but with significantly reduced training costs and much faster inference speeds. Our DiJiang-7B achieves comparable performance with LLaMA2-7B on various benchmark while requires only about 1/50 training cost. Code is available at https://github.com/YuchuanTian/DiJiang. |
https://proceedings.mlr.press/v235/chen24ac.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ac/chen24ac.pdf | https://openreview.net/forum?id=Y5Zi59N265 | GeoMFormer: A General Architecture for Geometric Molecular Representation Learning | https://proceedings.mlr.press/v235/chen24ac.html | Tianlang Chen, Shengjie Luo, Di He, Shuxin Zheng, Tie-Yan Liu, Liwei Wang | https://proceedings.mlr.press/v235/chen24ac.html | ICML 2024 | Molecular modeling, a central topic in quantum mechanics, aims to accurately calculate the properties and simulate the behaviors of molecular systems. The molecular model is governed by physical laws, which impose geometric constraints such as invariance and equivariance to coordinate rotation and translation. While numerous deep learning approaches have been developed to learn molecular representations under these constraints, most of them are built upon heuristic and costly modules. We argue that there is a strong need for a general and flexible framework for learning both invariant and equivariant features. In this work, we introduce a novel Transformer-based molecular model called GeoMFormer to achieve this goal. Using the standard Transformer modules, two separate streams are developed to maintain and learn invariant and equivariant representations. Carefully designed cross-attention modules bridge the two streams, allowing information fusion and enhancing geometric modeling in each stream. As a general and flexible architecture, we show that many previous architectures can be viewed as special instantiations of GeoMFormer. Extensive experiments are conducted to demonstrate the power of GeoMFormer. All empirical results show that GeoMFormer achieves strong performance on both invariant and equivariant tasks of different types and scales. Code and models will be made publicly available at https://github.com/c-tl/GeoMFormer. |
https://proceedings.mlr.press/v235/chen24ad.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ad/chen24ad.pdf | https://openreview.net/forum?id=v1I4zRAjMb | TENG: Time-Evolving Natural Gradient for Solving PDEs With Deep Neural Nets Toward Machine Precision | https://proceedings.mlr.press/v235/chen24ad.html | Zhuo Chen, Jacob Mccarran, Esteban Vizcaino, Marin Soljacic, Di Luo | https://proceedings.mlr.press/v235/chen24ad.html | ICML 2024 | Partial differential equations (PDEs) are instrumental for modeling dynamical systems in science and engineering. The advent of neural networks has initiated a significant shift in tackling these complexities though challenges in accuracy persist, especially for initial value problems. In this paper, we introduce the Time-Evolving Natural Gradient (TENG), generalizing time-dependent variational principles and optimization-based time integration, leveraging natural gradient optimization to obtain high accuracy in neural-network-based PDE solutions. Our comprehensive development includes algorithms like TENG-Euler and its high-order variants, such as TENG-Heun, tailored for enhanced precision and efficiency. TENG’s effectiveness is further validated through its performance, surpassing current leading methods and achieving machine precision in step-by-step optimizations across a spectrum of PDEs, including the heat equation, Allen-Cahn equation, and Burgers’ equation. |
https://proceedings.mlr.press/v235/chen24ae.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ae/chen24ae.pdf | https://openreview.net/forum?id=xFk0w9zoV3 | EE-LLM: Large-Scale Training and Inference of Early-Exit Large Language Models with 3D Parallelism | https://proceedings.mlr.press/v235/chen24ae.html | Yanxi Chen, Xuchen Pan, Yaliang Li, Bolin Ding, Jingren Zhou | https://proceedings.mlr.press/v235/chen24ae.html | ICML 2024 | We present EE-LLM, a framework for large-scale training and inference of early-exit large language models (LLMs). While recent works have shown preliminary evidence for the efficacy of early exiting in accelerating LLM inference, EE-LLM makes a foundational step towards scaling up early-exit LLMs by supporting their training and inference with massive 3D parallelism. Built upon Megatron-LM, EE-LLM implements a variety of algorithmic innovations and performance optimizations tailored to early exiting, including a lightweight method that facilitates backpropagation for the early-exit training objective with pipeline parallelism, techniques of leveraging idle resources in the original pipeline schedule for computation related to early-exit layers, and two approaches of early-exit inference that are compatible with KV caching for autoregressive generation. Our analytical and empirical study shows that EE-LLM achieves great training efficiency with negligible computational overhead compared to standard LLM training, as well as outstanding inference speedup without compromising output quality. To facilitate further research and adoption, we release EE-LLM at https://github.com/pan-x-c/EE-LLM. |
https://proceedings.mlr.press/v235/chen24af.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24af/chen24af.pdf | https://openreview.net/forum?id=AxmefV2NEf | TimeMIL: Advancing Multivariate Time Series Classification via a Time-aware Multiple Instance Learning | https://proceedings.mlr.press/v235/chen24af.html | Xiwen Chen, Peijie Qiu, Wenhui Zhu, Huayu Li, Hao Wang, Aristeidis Sotiras, Yalin Wang, Abolfazl Razi | https://proceedings.mlr.press/v235/chen24af.html | ICML 2024 | Deep neural networks, including transformers and convolutional neural networks (CNNs), have significantly improved multivariate time series classification (MTSC). However, these methods often rely on supervised learning, which does not fully account for the sparsity and locality of patterns in time series data (e.g., quantification of diseases-related anomalous points in ECG and abnormal detection in signal). To address this challenge, we formally discuss and reformulate MTSC as a weakly supervised problem, introducing a novel multiple-instance learning (MIL) framework for better localization of patterns of interest and modeling time dependencies within time series. Our novel approach, TimeMIL, formulates the temporal correlation and ordering within a time-aware MIL pooling, leveraging a tokenized transformer with a specialized learnable wavelet positional token. The proposed method surpassed 26 recent state-of-the-art MTSC methods, underscoring the effectiveness of the weakly supervised TimeMIL in MTSC. The code is available https://github.com/xiwenc1/TimeMIL. |
https://proceedings.mlr.press/v235/chen24ag.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ag/chen24ag.pdf | https://openreview.net/forum?id=PHUAG63Efe | AegisFL: Efficient and Flexible Privacy-Preserving Byzantine-Robust Cross-silo Federated Learning | https://proceedings.mlr.press/v235/chen24ag.html | Dong Chen, Hongyuan Qu, Guangwu Xu | https://proceedings.mlr.press/v235/chen24ag.html | ICML 2024 | Privacy attacks and poisoning attacks are two of the thorniest problems in federation learning (FL). Homomorphic encryption (HE), which allows certain mathematical operations to be done in the ciphertext state, provides a way to solve these two problems simultaneously. However, existing Paillier-based and CKKS-based privacy-preserving byzantine-robust FL (PBFL) solutions not only suffer from low efficiency but also expose the final model to the server. Additionally, these methods are limited to one robust aggregation algorithm (AGR) and are therefore vulnerable to AGR-tailored poisoning attacks. In this paper, we present AegisFL, an efficient PBLF system that provides the flexibility to change the AGR. We first observe that the core of the existing advanced AGRs is to calculate the inner products, $L_2$ norms and mean values for vectors. Based on this observation, we tailor a packing scheme for PBFL, which fits perfectly with RLWE-based fully homomorphic encryption. Under this packing scheme, the server only needs to perform one ciphertext multiplication to construct any required AGR, while the global model only belongs to honest clients. Finally, we conduct extensive experiments on different datasets and adversary settings, which also confirm the effectiveness and efficiency of our scheme. |
https://proceedings.mlr.press/v235/chen24ah.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ah/chen24ah.pdf | https://openreview.net/forum?id=ffLblkoCw8 | MAGDi: Structured Distillation of Multi-Agent Interaction Graphs Improves Reasoning in Smaller Language Models | https://proceedings.mlr.press/v235/chen24ah.html | Justin Chen, Swarnadeep Saha, Elias Stengel-Eskin, Mohit Bansal | https://proceedings.mlr.press/v235/chen24ah.html | ICML 2024 | Multi-agent interactions between Large Language Model (LLM) agents have shown major improvements on diverse reasoning tasks. However, these involve long generations from multiple models across several rounds, making them expensive. Moreover, these multi-agent approaches fail to provide a final, single model for efficient inference. To address this, we introduce MAGDi, a new method for structured distillation of the reasoning interactions between multiple LLMs into smaller LMs. MAGDi teaches smaller models by representing multi-agent interactions as graphs, augmenting a base student model with a graph encoder, and distilling knowledge using three objective functions: next-token prediction, a contrastive loss between correct and incorrect reasoning, and a graph-based objective to model the interaction structure. Experiments on seven widely used commonsense and math reasoning benchmarks show that MAGDi improves the reasoning capabilities of smaller models, outperforming several methods that distill from a single teacher and multiple teachers. Moreover, MAGDi also demonstrates an order of magnitude higher efficiency over its teachers. We conduct extensive analyses to show that MAGDi (1) enhances the generalizability to out-of-domain tasks, (2) scales positively with the size and strength of the base student model, and (3) obtains larger improvements (via our multi-teacher training) when applying self-consistency – an inference technique that relies on model diversity. |
https://proceedings.mlr.press/v235/chen24ai.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ai/chen24ai.pdf | https://openreview.net/forum?id=sLZzFTMWSt | CaRiNG: Learning Temporal Causal Representation under Non-Invertible Generation Process | https://proceedings.mlr.press/v235/chen24ai.html | Guangyi Chen, Yifan Shen, Zhenhao Chen, Xiangchen Song, Yuewen Sun, Weiran Yao, Xiao Liu, Kun Zhang | https://proceedings.mlr.press/v235/chen24ai.html | ICML 2024 | Identifying the underlying time-delayed latent causal processes in sequential data is vital for grasping temporal dynamics and making downstream reasoning. While some recent methods can robustly identify these latent causal variables, they rely on strict assumptions about the invertible generation process from latent variables to observed data. However, these assumptions are often hard to satisfy in real-world applications containing information loss. For instance, the visual perception process translates a 3D space into 2D images, or the phenomenon of persistence of vision incorporates historical data into current perceptions. To address this challenge, we establish an identifiability theory that allows for the recovery of independent latent components even when they come from a nonlinear and non-invertible mix. Using this theory as a foundation, we propose a principled approach, CaRiNG, to learn the Causal Representation of Non-invertible Generative temporal data with identifiability guarantees. Specifically, we utilize temporal context to recover lost latent information and apply the conditions in our theory to guide the training process. Through experiments conducted on synthetic datasets, we validate that our CaRiNG method reliably identifies the causal process, even when the generation process is non-invertible. Moreover, we demonstrate that our approach considerably improves temporal understanding and reasoning in practical applications. |
https://proceedings.mlr.press/v235/chen24aj.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24aj/chen24aj.pdf | https://openreview.net/forum?id=d2f2sCXQuI | GRATH: Gradual Self-Truthifying for Large Language Models | https://proceedings.mlr.press/v235/chen24aj.html | Weixin Chen, Dawn Song, Bo Li | https://proceedings.mlr.press/v235/chen24aj.html | ICML 2024 | Truthfulness is paramount for large language models (LLMs) as they are increasingly deployed in real-world applications. However, existing LLMs still struggle with generating truthful content, as evidenced by their modest performance on benchmarks like TruthfulQA. To address this issue, we propose GRAdual self-truTHifying (GRATH), a novel post-processing method to enhance truthfulness of LLMs. GRATH utilizes out-of-domain question prompts to generate pairwise truthfulness training data with each pair containing a question and its correct and incorrect answers, and then optimizes the model via direct preference optimization (DPO) to learn from the truthfulness difference between answer pairs. GRATH iteratively refines truthfulness data and updates the model, leading to a gradual improvement in model truthfulness in a self-supervised manner. Empirically, we evaluate GRATH using different 7B-LLMs and compare with LLMs with similar or even larger sizes on benchmark datasets. Our results show that GRATH effectively improves LLMs’ truthfulness without compromising other core capabilities. Notably, GRATH achieves state-of-the-art performance on TruthfulQA, with MC1 accuracy of 54.71% and MC2 accuracy of 69.10%, which even surpass those on 70B-LLMs. The code is available at https://github.com/chenweixin107/GRATH. |
https://proceedings.mlr.press/v235/chen24ak.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ak/chen24ak.pdf | https://openreview.net/forum?id=QhHMx51ir6 | Unleashing the Power of Meta-tuning for Few-shot Generalization Through Sparse Interpolated Experts | https://proceedings.mlr.press/v235/chen24ak.html | Shengzhuang Chen, Jihoon Tack, Yunqiao Yang, Yee Whye Teh, Jonathan Richard Schwarz, Ying Wei | https://proceedings.mlr.press/v235/chen24ak.html | ICML 2024 | Recent successes suggest that parameter-efficient fine-tuning of foundation models is becoming the state-of-the-art method for transfer learning in vision, gradually replacing the rich literature of alternatives such as meta-learning. In trying to harness the best of both worlds, meta-tuning introduces a subsequent optimization stage of foundation models but has so far only shown limited success and crucially tends to underperform on out-of-distribution (OOD) tasks. In this paper, we introduce Sparse MetA-Tuning (SMAT), a method inspired by sparse mixture-of-experts approaches and trained to isolate subsets of pre-trained parameters automatically for meta-tuning on each task. SMAT successfully overcomes OOD sensitivity and delivers on the promise of enhancing the transfer abilities of vision foundation models beyond parameter-efficient finetuning. We establish new state-of-the-art results on a challenging combination of Meta-Dataset augmented with additional OOD tasks in both zero-shot and gradient-based adaptation settings. In addition, we provide a thorough analysis of the superiority of learned over hand-designed sparsity patterns for sparse expert methods and the pivotal importance of the sparsity level in balancing between in-distribution and out-of-distribution generalization. Our code and models are publicly available. |
https://proceedings.mlr.press/v235/chen24al.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24al/chen24al.pdf | https://openreview.net/forum?id=yHs3jIPgaF | Performative Prediction with Bandit Feedback: Learning through Reparameterization | https://proceedings.mlr.press/v235/chen24al.html | Yatong Chen, Wei Tang, Chien-Ju Ho, Yang Liu | https://proceedings.mlr.press/v235/chen24al.html | ICML 2024 | Performative prediction, as introduced by Perdomo et al., is a framework for studying social prediction in which the data distribution itself changes in response to the deployment of a model. Existing work in this field usually hinges on three assumptions that are easily violated in practice: that the performative risk is convex over the deployed model, that the mapping from the model to the data distribution is known to the model designer in advance, and the first-order information of the performative risk is available. In this paper, we initiate the study of performative prediction problems that do not require these assumptions. Specifically, we develop a parameterization framework that parametrizes the performative prediction objective as a function of the induced data distribution. We also develop a two-level zeroth-order optimization procedure, where the first level performs iterative optimization on the distribution parameter space, and the second level learns the model that induced a particular target distribution parameter at each iteration. Under mild conditions, this reparameterization allows us to transform the non-convex objective into a convex one and achieve provable regret guarantees. In particular, we provide a regret bound that is sublinear in the total number of performative samples taken and is only polynomial in the dimension of the model parameter. |
https://proceedings.mlr.press/v235/chen24am.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24am/chen24am.pdf | https://openreview.net/forum?id=4RqG4K5UwL | Self-Attention through Kernel-Eigen Pair Sparse Variational Gaussian Processes | https://proceedings.mlr.press/v235/chen24am.html | Yingyi Chen, Qinghua Tao, Francesco Tonin, Johan Suykens | https://proceedings.mlr.press/v235/chen24am.html | ICML 2024 | While the great capability of Transformers significantly boosts prediction accuracy, it could also yield overconfident predictions and require calibrated uncertainty estimation, which can be commonly tackled by Gaussian processes (GPs). Existing works apply GPs with symmetric kernels under variational inference to the attention kernel; however, omitting the fact that attention kernels are in essence asymmetric. Moreover, the complexity of deriving the GP posteriors remains high for large-scale data. In this work, we propose Kernel-Eigen Pair Sparse Variational Gaussian Processes (KEP-SVGP) for building uncertainty-aware self-attention where the asymmetry of attention kernels is tackled by Kernel SVD (KSVD) and a reduced complexity is acquired. Through KEP-SVGP, i) the SVGP pair induced by the two sets of singular vectors from KSVD w.r.t. the attention kernel fully characterizes the asymmetry; ii) using only a small set of adjoint eigenfunctions from KSVD, the derivation of SVGP posteriors can be based on the inversion of a diagonal matrix containing singular values, contributing to a reduction in time complexity; iii) an evidence lower bound is derived so that variational parameters and network weights can be optimized with it. Experiments verify our excellent performances and efficiency on in-distribution, distribution-shift and out-of-distribution benchmarks. |
https://proceedings.mlr.press/v235/chen24an.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24an/chen24an.pdf | https://openreview.net/forum?id=E41gvBG4s6 | Recovering Labels from Local Updates in Federated Learning | https://proceedings.mlr.press/v235/chen24an.html | Huancheng Chen, Haris Vikalo | https://proceedings.mlr.press/v235/chen24an.html | ICML 2024 | Gradient inversion (GI) attacks present a threat to the privacy of clients in federated learning (FL) by aiming to enable reconstruction of the clients’ data from communicated model updates. A number of such techniques attempts to accelerate data recovery by first reconstructing labels of the samples used in local training. However, existing label extraction methods make strong assumptions that typically do not hold in realistic FL settings. In this paper we present a novel label recovery scheme, Recovering Labels from Local Updates (RLU), which provides near-perfect accuracy when attacking untrained (most vulnerable) models. More significantly, RLU achieves high performance even in realistic real-world settings where the clients in an FL system run multiple local epochs, train on heterogeneous data, and deploy various optimizers to minimize different objective functions. Specifically, RLU estimates labels by solving a least-square problem that emerges from the analysis of the correlation between labels of the data points used in a training round and the resulting update of the output layer. The experimental results on several datasets, architectures, and data heterogeneity scenarios demonstrate that the proposed method consistently outperforms existing baselines, and helps improve quality of the reconstructed images in GI attacks in terms of both PSNR and LPIPS. |
https://proceedings.mlr.press/v235/chen24ao.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ao/chen24ao.pdf | https://openreview.net/forum?id=gjgRKbdYR7 | SelfIE: Self-Interpretation of Large Language Model Embeddings | https://proceedings.mlr.press/v235/chen24ao.html | Haozhe Chen, Carl Vondrick, Chengzhi Mao | https://proceedings.mlr.press/v235/chen24ao.html | ICML 2024 | How do large language models (LLMs) obtain their answers? The ability to explain and control an LLM’s reasoning process is key for reliability, transparency, and future model developments. We propose SelfIE (Self-Interpretation of Embeddings), a framework that enables LLMs to interpret their own embeddings in natural language by leveraging their ability to respond to inquiries about a given passage. Capable of interpreting open-world concepts in the hidden embeddings, SelfIE reveals LLM internal reasoning in cases such as making ethical decisions, internalizing prompt injection, and recalling harmful knowledge. SelfIE’s text descriptions on hidden embeddings open avenues to control LLM reasoning. We propose Supervised Control, which allows editing open-ended concepts while only requiring gradient computation of individual layer. We extend RLHF to hidden embeddings and propose Reinforcement Control that erases harmful knowledge in LLM without supervision targets. |
https://proceedings.mlr.press/v235/chen24ap.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ap/chen24ap.pdf | https://openreview.net/forum?id=RuH78kOcDi | Locally Differentially Private Decentralized Stochastic Bilevel Optimization with Guaranteed Convergence Accuracy | https://proceedings.mlr.press/v235/chen24ap.html | Ziqin Chen, Yongqiang Wang | https://proceedings.mlr.press/v235/chen24ap.html | ICML 2024 | Decentralized bilevel optimization based machine learning techniques are achieving remarkable success in a wide variety of domains. However, the intensive exchange of information (involving nested-loops of consensus or communication iterations) in existing decentralized bilevel optimization algorithms leads to a great challenge to ensure rigorous differential privacy, which, however, is necessary to bring the benefits of machine learning to domains where involved data are sensitive. By proposing a new decentralized stochastic bilevel-optimization algorithm which avoids nested-loops of information-exchange iterations, we achieve, for the first time, both differential privacy and accurate convergence in decentralized bilevel optimization. This is significant since even for single-level decentralized optimization and learning, existing differential-privacy solutions have to sacrifice convergence accuracy for privacy. Besides characterizing the convergence rate under nonconvex/convex/strongly convex conditions, we also rigorously quantify the price of differential privacy in the convergence rate. Experimental results on machine learning models confirm the efficacy of our algorithm. |
https://proceedings.mlr.press/v235/chen24aq.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24aq/chen24aq.pdf | https://openreview.net/forum?id=hQpUhySEJi | Subequivariant Reinforcement Learning in 3D Multi-Entity Physical Environments | https://proceedings.mlr.press/v235/chen24aq.html | Runfa Chen, Ling Wang, Yu Du, Tianrui Xue, Fuchun Sun, Jianwei Zhang, Wenbing Huang | https://proceedings.mlr.press/v235/chen24aq.html | ICML 2024 | Learning policies for multi-entity systems in 3D environments is far more complicated against single-entity scenarios, due to the exponential expansion of the global state space as the number of entities increases. One potential solution of alleviating the exponential complexity is dividing the global space into independent local views that are invariant to transformations including translations and rotations. To this end, this paper proposes Subequivariant Hierarchical Neural Networks (SHNN) to facilitate multi-entity policy learning. In particular, SHNN first dynamically decouples the global space into local entity-level graphs via task assignment. Second, it leverages subequivariant message passing over the local entity-level graphs to devise local reference frames, remarkably compressing the representation redundancy, particularly in gravity-affected environments. Furthermore, to overcome the limitations of existing benchmarks in capturing the subtleties of multi-entity systems under the Euclidean symmetry, we propose the Multi-entity Benchmark (MEBEN), a new suite of environments tailored for exploring a wide range of multi-entity reinforcement learning. Extensive experiments demonstrate significant advancements of SHNN on the proposed benchmarks compared to existing methods. Comprehensive ablations are conducted to verify the indispensability of task assignment and subequivariance. |
https://proceedings.mlr.press/v235/chen24ar.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ar/chen24ar.pdf | https://openreview.net/forum?id=7sgqXa4aNM | A General Framework for Learning from Weak Supervision | https://proceedings.mlr.press/v235/chen24ar.html | Hao Chen, Jindong Wang, Lei Feng, Xiang Li, Yidong Wang, Xing Xie, Masashi Sugiyama, Rita Singh, Bhiksha Raj | https://proceedings.mlr.press/v235/chen24ar.html | ICML 2024 | Weakly supervised learning generally faces challenges in applicability to various scenarios with diverse weak supervision and in scalability due to the complexity of existing algorithms, thereby hindering the practical deployment. This paper introduces a general framework for learning from weak supervision (GLWS) with a novel algorithm. Central to GLWS is an Expectation-Maximization (EM) formulation, adeptly accommodating various weak supervision sources, including instance partial labels, aggregate statistics, pairwise observations, and unlabeled data. We further present an advanced algorithm that significantly simplifies the EM computational demands using a Non-deterministic Finite Automaton (NFA) along with a forward-backward algorithm, which effectively reduces time complexity from quadratic or factorial often required in existing solutions to linear scale. The problem of learning from arbitrary weak supervision is therefore converted to the NFA modeling of them. GLWS not only enhances the scalability of machine learning models but also demonstrates superior performance and versatility across 11 weak supervision scenarios. We hope our work paves the way for further advancements and practical deployment in this field. |
https://proceedings.mlr.press/v235/chen24as.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24as/chen24as.pdf | https://openreview.net/forum?id=OnidGtOhg3 | Diffusion Model-Augmented Behavioral Cloning | https://proceedings.mlr.press/v235/chen24as.html | Shang-Fu Chen, Hsiang-Chun Wang, Ming-Hao Hsu, Chun-Mao Lai, Shao-Hua Sun | https://proceedings.mlr.press/v235/chen24as.html | ICML 2024 | Imitation learning addresses the challenge of learning by observing an expert’s demonstrations without access to reward signals from environments. Most existing imitation learning methods that do not require interacting with environments either model the expert distribution as the conditional probability p(a|s) (e.g., behavioral cloning, BC) or the joint probability p(s, a). Despite the simplicity of modeling the conditional probability with BC, it usually struggles with generalization. While modeling the joint probability can improve generalization performance, the inference procedure is often time-consuming, and the model can suffer from manifold overfitting. This work proposes an imitation learning framework that benefits from modeling both the conditional and joint probability of the expert distribution. Our proposed Diffusion Model-Augmented Behavioral Cloning (DBC) employs a diffusion model trained to model expert behaviors and learns a policy to optimize both the BC loss (conditional) and our proposed diffusion model loss (joint). DBC outperforms baselines in various continuous control tasks in navigation, robot arm manipulation, dexterous manipulation, and locomotion. We design additional experiments to verify the limitations of modeling either the conditional probability or the joint probability of the expert distribution, as well as compare different generative models. Ablation studies justify the effectiveness of our design choices. |
https://proceedings.mlr.press/v235/chen24at.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24at/chen24at.pdf | https://openreview.net/forum?id=Rp8R9C0Sth | AutoOS: Make Your OS More Powerful by Exploiting Large Language Models | https://proceedings.mlr.press/v235/chen24at.html | Huilai Chen, Yuanbo Wen, Limin Cheng, Shouxu Kuang, Yumeng Liu, Weijia Li, Ling Li, Rui Zhang, Xinkai Song, Wei Li, Qi Guo, Yunji Chen | https://proceedings.mlr.press/v235/chen24at.html | ICML 2024 | With the rapid development of Artificial Intelligence of Things (AIoT), customizing and optimizing operating system (OS) kernel configurations for various AIoT application scenarios is crucial for maximizing system performance. However, existing approaches falter due to the overwhelming problem complexity (i.e., over 15,000 configuration options in the Linux kernel), together with the huge evaluation costs and error-prone options that may result in OS boot-up failure, which all make it an unresolved problem to optimize the Linux kernel automatically. In this paper, we introduce AutoOS, a novel framework exploiting Large Language Models for customizing and optimizing OS kernel configurations automatically for various AIoT application scenarios.Inspired by the inherently directory-structured kernel configuration process, we first formulate our research problem as optimizing on a dynamic tree. We then propose a novel framework integrating a state machine-based traversal algorithm as the observe-prune-propose-act-correct loop, which can effectively refine the optimization space and ensure a successful OS boot-up.Experimental results show that AutoOS can automatically customize and optimize the OS kernel configurations without human effort. More importantly, AutoOS even achieves better performance by up to 25% than vendor-provided configuration. |
https://proceedings.mlr.press/v235/chen24au.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24au/chen24au.pdf | https://openreview.net/forum?id=VOcsmIBiXE | Positional Knowledge is All You Need: Position-induced Transformer (PiT) for Operator Learning | https://proceedings.mlr.press/v235/chen24au.html | Junfeng Chen, Kailiang Wu | https://proceedings.mlr.press/v235/chen24au.html | ICML 2024 | Operator learning for Partial Differential Equations (PDEs) is rapidly emerging as a promising approach for surrogate modeling of intricate systems. Transformers with the self-attention mechanism—a powerful tool originally designed for natural language processing—have recently been adapted for operator learning. However, they confront challenges, including high computational demands and limited interpretability. This raises a critical question: Is there a more efficient attention mechanism for Transformer-based operator learning? This paper proposes the Position-induced Transformer (PiT), built on an innovative position-attention mechanism, which demonstrates significant advantages over the classical self-attention in operator learning. Position-attention draws inspiration from numerical methods for PDEs. Different from self-attention, position-attention is induced by only the spatial interrelations of sampling positions for input functions of the operators, and does not rely on the input function values themselves, thereby greatly boosting efficiency. PiT exhibits superior performance over current state-of-the-art neural operators in a variety of complex operator learning tasks across diverse PDE benchmarks. Additionally, PiT possesses an enhanced discretization convergence feature, compared to the widely-used Fourier neural operator. |
https://proceedings.mlr.press/v235/chen24av.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24av/chen24av.pdf | https://openreview.net/forum?id=s3e8poX3kb | In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation | https://proceedings.mlr.press/v235/chen24av.html | Shiqi Chen, Miao Xiong, Junteng Liu, Zhengxuan Wu, Teng Xiao, Siyang Gao, Junxian He | https://proceedings.mlr.press/v235/chen24av.html | ICML 2024 | Large language models (LLMs) frequently hallucinate, e.g., making factual errors, yet our understanding of why they make these errors remains limited. In this study, we aim to understand the underlying mechanisms of LLM hallucinations from the perspective of inner representations. We discover a pattern associated with hallucinations: correct generations tend to have sharper context activations in the hidden states of the in-context tokens, compared to that of the incorrect generations. Leveraging this signal, we propose an entropy-based metric to quantify the sharpness among the in-context hidden states and incorporate it into the decoding process, i.e, use the entropy value to adjust the next token prediction distribution to improve the factuality and overall quality of the generated text. Experiments on knowledge-seeking datasets (Natural Questions, HotpotQA, TriviaQA) and hallucination benchmark (TruthfulQA) demonstrate our consistent effectiveness, e.g., up to 8.6 absolute points on TruthfulQA. We believe this study can improve our understanding of hallucinations and serve as a practical solution for hallucination mitigation. |
https://proceedings.mlr.press/v235/chen24aw.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24aw/chen24aw.pdf | https://openreview.net/forum?id=pQyoBWA146 | Split-Ensemble: Efficient OOD-aware Ensemble via Task and Model Splitting | https://proceedings.mlr.press/v235/chen24aw.html | Anthony Chen, Huanrui Yang, Yulu Gan, Denis A Gudovskiy, Zhen Dong, Haofan Wang, Tomoyuki Okuno, Yohei Nakata, Kurt Keutzer, Shanghang Zhang | https://proceedings.mlr.press/v235/chen24aw.html | ICML 2024 | Uncertainty estimation is crucial for deep learning models to detect out-of-distribution (OOD) inputs. However, the naive deep learning classifiers produce uncalibrated uncertainty for OOD data. Improving the uncertainty estimation typically requires external data for OOD-aware training or considerable costs to build an ensemble. In this work, we improve on uncertainty estimation without extra OOD data or additional inference costs using an alternative Split-Ensemble method. Specifically, we propose a novel subtask-splitting ensemble training objective where a task is split into several complementary subtasks based on feature similarity. Each subtask considers part of the data as in distribution while all the rest as OOD data. Diverse submodels can therefore be trained on each subtask with OOD-aware objectives, learning generalizable uncertainty estimation. To avoid overheads, we enable low-level feature sharing among submodels, building a tree-like Split-Ensemble architecture via iterative splitting and pruning. Empirical study shows Split-Ensemble, without additional computational cost, improves accuracy over a single model by 0.8%, 1.8%, and 25.5% on CIFAR-10, CIFAR-100, and Tiny-ImageNet, respectively. OOD detection for the same backbone and in-distribution datasets surpasses a single model baseline by 2.2%, 8.1%, and 29.6% in mean AUROC, respectively. |
https://proceedings.mlr.press/v235/chen24ax.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ax/chen24ax.pdf | https://openreview.net/forum?id=DJdVzxemdA | Deep Demonstration Tracing: Learning Generalizable Imitator Policy for Runtime Imitation from a Single Demonstration | https://proceedings.mlr.press/v235/chen24ax.html | Xiong-Hui Chen, Junyin Ye, Hang Zhao, Yi-Chen Li, Xu-Hui Liu, Haoran Shi, Yu-Yan Xu, Zhihao Ye, Si-Hang Yang, Yang Yu, Anqi Huang, Kai Xu, Zongzhang Zhang | https://proceedings.mlr.press/v235/chen24ax.html | ICML 2024 | One-shot imitation learning (OSIL) is to learn an imitator agent that can execute multiple tasks with only a single demonstration. In real-world scenario, the environment is dynamic, e.g., unexpected changes can occur after demonstration. Thus, achieving generalization of the imitator agent is crucial as agents would inevitably face situations unseen in the provided demonstrations. While traditional OSIL methods excel in relatively stationary settings, their adaptability to such unforeseen changes, which asking for a higher level of generalization ability for the imitator agents, is limited and rarely discussed. In this work, we present a new algorithm called Deep Demonstration Tracing (DDT). In DDT, we propose a demonstration transformer architecture to encourage agents to adaptively trace suitable states in demonstrations. Besides, it integrates OSIL into a meta-reinforcement-learning training paradigm, providing regularization for policies in unexpected situations. We evaluate DDT on a new navigation task suite and robotics tasks, demonstrating its superior performance over existing OSIL methods across all evaluated tasks in dynamic environments with unforeseen changes. The project page is in https://osil-ddt.github.io. |
https://proceedings.mlr.press/v235/chen24ay.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ay/chen24ay.pdf | https://openreview.net/forum?id=oRLwyayrh1 | DRCT: Diffusion Reconstruction Contrastive Training towards Universal Detection of Diffusion Generated Images | https://proceedings.mlr.press/v235/chen24ay.html | Baoying Chen, Jishen Zeng, Jianquan Yang, Rui Yang | https://proceedings.mlr.press/v235/chen24ay.html | ICML 2024 | Diffusion models have made significant strides in visual content generation but also raised increasing demands on generated image detection. Existing detection methods have achieved considerable progress, but they usually suffer a significant decline in accuracy when detecting images generated by an unseen diffusion model. In this paper, we seek to address the generalizability of generated image detectors from the perspective of hard sample classification. The basic idea is that if a classifier can distinguish generated images that closely resemble real ones, then it can also effectively detect less similar samples, potentially even those produced by a different diffusion model. Based on this idea, we propose Diffusion Reconstruction Contrastive Learning (DRCT), a universal framework to enhance the generalizability of the existing detectors. DRCT generates hard samples by high-quality diffusion reconstruction and adopts contrastive training to guide the learning of diffusion artifacts. In addition, we have built a million-scale dataset, DRCT-2M, including 16 types diffusion models for the evaluation of generalizability of detection methods. Extensive experimental results show that detectors enhanced with DRCT achieve over a 10% accuracy improvement in cross-set tests. The code, models, and dataset will soon be available at https://github.com/beibuwandeluori/DRCT. |
https://proceedings.mlr.press/v235/chen24az.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24az/chen24az.pdf | https://openreview.net/forum?id=yShA4VPYZB | $\rm E(3)$-Equivariant Actor-Critic Methods for Cooperative Multi-Agent Reinforcement Learning | https://proceedings.mlr.press/v235/chen24az.html | Dingyang Chen, Qi Zhang | https://proceedings.mlr.press/v235/chen24az.html | ICML 2024 | Identification and analysis of symmetrical patterns in the natural world have led to significant discoveries across various scientific fields, such as the formulation of gravitational laws in physics and advancements in the study of chemical structures. In this paper, we focus on exploiting Euclidean symmetries inherent in certain cooperative multi-agent reinforcement learning (MARL) problems and prevalent in many applications. We begin by formally characterizing a subclass of Markov games with a general notion of symmetries that admits the existence of symmetric optimal values and policies. Motivated by these properties, we design neural network architectures with symmetric constraints embedded as an inductive bias for multi-agent actor-critic methods. This inductive bias results in superior performance in various cooperative MARL benchmarks and impressive generalization capabilities such as zero-shot learning and transfer learning in unseen scenarios with repeated symmetric patterns. |
https://proceedings.mlr.press/v235/chen24ba.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24ba/chen24ba.pdf | https://openreview.net/forum?id=jrHUbftLd6 | FedMBridge: Bridgeable Multimodal Federated Learning | https://proceedings.mlr.press/v235/chen24ba.html | Jiayi Chen, Aidong Zhang | https://proceedings.mlr.press/v235/chen24ba.html | ICML 2024 | Multimodal Federated Learning (MFL) addresses the setup of multiple clients with diversified modality types (e.g. image, text, video, and audio) working together to improve their local personal models in a data-privacy manner. Prior MFL works rely on restrictive compositional neural architecture designs to ensure inter-client information sharing via blockwise model aggregation, limiting their applicability in the real-world Architecture-personalized MFL (AMFL) scenarios, where clients may have distinguished multimodal interaction strategies and there is no restriction on local architecture design. The key challenge in AMFL is how to automatically and efficiently tackle the two heterogeneity patterns–statistical and architecture heterogeneity–while maximizing the beneficial information sharing among clients. To solve this challenge, we propose FedMBridge, which leverages a topology-aware hypernetwork to act as a bridge that can automatically balance and digest the two heterogeneity patterns in a communication-efficient manner. Our experiments on four AMFL simulations demonstrate the efficiency and effectiveness of our proposed approach. |
https://proceedings.mlr.press/v235/chen24bb.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bb/chen24bb.pdf | https://openreview.net/forum?id=rkYOxLLv2x | Revealing the Dark Secrets of Extremely Large Kernel ConvNets on Robustness | https://proceedings.mlr.press/v235/chen24bb.html | Honghao Chen, Yurong Zhang, Xiaokun Feng, Xiangxiang Chu, Kaiqi Huang | https://proceedings.mlr.press/v235/chen24bb.html | ICML 2024 | Robustness is a vital aspect to consider when deploying deep learning models into the wild. Numerous studies have been dedicated to the study of the robustness of vision transformers (ViTs), which have dominated as the mainstream backbone choice for vision tasks since the dawn of 2020s. Recently, some large kernel convnets make a comeback with impressive performance and efficiency. However, it still remains unclear whether large kernel networks are robust and the attribution of their robustness. In this paper, we first conduct a comprehensive evaluation of large kernel convnets’ robustness and their differences from typical small kernel counterparts and ViTs on six diverse robustness benchmark datasets. Then to analyze the underlying factors behind their strong robustness, we design experiments from both quantitative and qualitative perspectives to reveal large kernel convnets’ intriguing properties that are completely different from typical convnets. Our experiments demonstrate for the first time that pure CNNs can achieve exceptional robustness comparable or even superior to that of ViTs. Our analysis on occlusion invariance, kernel attention patterns and frequency characteristics provide novel insights into the source of robustness. Code available at: https://github.com/Lauch1ng/LKRobust. |