---
title: GameTheory Chat
emoji: 🎯
colorFrom: green
colorTo: green
sdk: gradio
sdk_version: 5.12.0
app_file: app.py
hardware: zero-a10g
models:
  - Qwen/Qwen2.5-7B-Instruct
  - Alogotron/GameTheory-Formulator-Model
  - Alogotron/GameTheory-Solver
datasets:
  - Alogotron/GameTheory-Bench
  - Alogotron/GameTheory-Formulator
tags:
  - game-theory
  - gradio
  - chat
  - qwen
  - lora
  - zerogpu
pinned: true
license: apache-2.0
short_description: AI-powered game theory formulation and solving
---

# 🎯 GameTheory Chat — AI-Powered Strategic Reasoning

Chat with a fine-tuned game theory AI. Describe real-world strategic scenarios and get formal game-theoretic analysis, or paste formal problems for direct solving.

## Features

- **Strategy Chat**: Describe any real-world strategic scenario → AI formulates it as a game theory model and solves it
- **Quick Solve**: Paste formal game theory problems (payoff matrices, etc.) → get step-by-step solutions
- **Streaming**: Real-time token-by-token response generation

## Models

| Model | Purpose | Training |
|-------|---------|----------|
| [GameTheory-Formulator-Model](https://huggingface.co/Alogotron/GameTheory-Formulator-Model) | Real-world → formal game theory | SFT on 1,215 formulation problems |
| [GameTheory-Solver](https://huggingface.co/Alogotron/GameTheory-Solver) | Solve formal game theory problems | SFT on 2,913 verified problems |

Base model: [Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) with QLoRA adapters (r=32, α=64).

## Part of the GameTheory-Bench Project

A 3-phase fine-tuning pipeline achieving 94% accuracy on game theory benchmarks.
See the [full technical blog](https://huggingface.co/spaces/Alogotron/game-theory-llm-blog) for details.