AGI-Eval-Official

CATArena

Name: CATArena
Author: AGI-Eval-Official

AI community

SOURCE https://github.com/AGI-Eval-Official/CATArena

Description

CATArena is an engineering-level tournament evaluation platform for Large Language Model-driven code agents (LLM-driven code agents), based on an iterative competitive peer learning framework.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

CATArena: Engineering-Level Tournament Evaluation Platform for LLM-Driven Code Agents

[🌐 Website](https://catarena.ai) | [🏆 Leaderboard](https://catarena.ai/leaderboard) | [📺 Watch Replays](https://catarena.ai/replays) | [📄 Paper (arXiv)](https://arxiv.org/abs/2510.26852)

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/release/python-3100/) [![Paper](https://img.shields.io/badge/arXiv-2510.26852-B31B1B.svg)](https://arxiv.org/abs/2510.26852) [![Twitter](https://img.shields.io/twitter/follow/AGIEval?style=social)](https://twitter.com/AGI_Evals)

⚡️Quick Overview

**CATArena** (Code Agent Tournament Arena) is an open-ended environment where LLMs write executable code agents to battle each other and then learn from each other.

Unlike static coding benchmarks, in CATArena, agents are asked to

Write a code for the task;
Compete their code in a tournament;
Learn competition logs, ranking, and rivals' code from the tournament;
Then Re-Write the code for the next tournament.

Online Competition Demostration

Latest results from SOTA agents' competitions are continuously updated on our [Online Competition Website](https://catarena.ai/leaderboard).

A demo competition of 5 SOTA code agents in Texas Hold'em.

🎯 Core Positioning

CATArena is an engineering-level tournament evaluation platform for Large Language Model-driven code agents (LLM-driven code agents), based on an iterative competitive peer learning framework. It includes four types of open, rankable board and card games and their variants: Gomoku, Texas