Tools and prompt templates used to build and evaluate SWE-rebench-v2 tasks for the paper.

SkyRL: A Modular Full-stack RL Library for LLMs

A version of verl to support diverse tool use [TMLR 2026]

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, …

Repository-level QA benchmark for software engineering LLMs

Rethinking Code Editing for Efficient Software Engineering Agents

The agent that grows with you

Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.

Extracted system prompts from Anthropic - Claude Fable 5, Opus 5, Claude Design, Claude Code. OpenAI - ChatGPT GPT-5.6-Sol, Codex. Google - Gemini 3.5 Flash, 3.1 Pro, Antigravity. xAI - Grok, Curso…

👩‍⚖️ Agent-as-a-Judge: The Magic for Open-Endedness

The first real AI developer

open source SWE-Atlas

TDD-Bench-Verified is a new benchmark for generating test cases for test-driven development (TDD)

Exercism exercises in Python.

aider is AI pair programming in your terminal

Agentless🐱: an agentless approach to automatically solve software development problems

Qwen3-Coder is the code version of Qwen3, the large language model series developed by Qwen team.

A conda-forge distribution.

Supercharge Your LLM Application Evaluations 🚀

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

PyTorch implementation of soft actor critic

An alignment auditing agent capable of quickly exploring alignment hypothesis

Realistic examples of building evals and optimizing agents with Harbor