facebookresearch/aira-dojo

AIRS-Bench: an AI Research Science benchmark for quantifying the end-to-end AI research abilities of LLM agents

Research infra for creating RL environments, post-training, and evals.

Creating realistic RL environments requires collaboration between researchers, engineers, and domain experts across many dimensions: artifacts, environments tools, dynamism of the environment, reproducibility, and more. There is no open source framework for building these environments effectively. Until now.

Framework for evaluating and improving agents

Official implement on InternBootCamp

Evaluate and improve models and agents using environments

Scalable RL for Any Agent and Sandbox.