THUSI-Lab/GameVerse

GameVerse: Can Vision-Language Models Learn from Video-based Reflection?

32
/ 100
Emerging

This tool is a comprehensive platform for evaluating how well AI models, specifically Vision-Language Models (VLMs), can play various video games. It takes game video footage and human-like strategic prompts as input and outputs performance metrics and behavioral analyses of the AI agent's gameplay. It's designed for AI researchers and game AI developers who are building and testing advanced game-playing agents.

Use this if you need to systematically benchmark the game-playing capabilities of your AI agents or Vision-Language Models across a wide array of complex video games.

Not ideal if you're a casual gamer looking for an AI opponent, or if you only need to evaluate a single, simple game with basic metrics.

game-AI-evaluation vision-language-models AI-benchmarking game-agent-development machine-gaming
No Package No Dependents
Maintenance 13 / 25
Adoption 8 / 25
Maturity 11 / 25
Community 0 / 25

How are scores calculated?

Stars

45

Forks

Language

Python

License

MIT

Last pushed

Mar 26, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/THUSI-Lab/GameVerse"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.