---
title: Beam Search Demo
emoji: 🌍
colorFrom: green
colorTo: indigo
sdk: gradio
sdk_version: 5.16.2
app_file: app.py
pinned: false
license: mit
short_description: Beam Search vs Greedy Search Demo
---

Welcome to the Beam Search vs Greedy Search Demo repository! This educational demo illustrates two text generation strategies in LLM inference using the Qwen2.5-0.5B model.

This demo compares:
- **Greedy Search**: Picks the most probable token at every generation step (deterministic).
- **Beam Search**: Explores multiple beams concurrently and returns the top candidate, often achieving more coherent or creative outputs.