rl-environments 开源项目推荐（68 个）

harbor-framework /harbor

评估与改进智能体的框架

benchflow-ai /awesome-evals

一个精心策划的非BS图书馆，包含构建和评估人工智能代理的最佳资源--论文、博客、讲座、工具、基准。由BenchFlow维护。

huggingface /Repo2RLEnv

Turn any repository into verifiable RL environments for coding agents - Harbor tasks you can train on, evaluate and share on the Hugging Face Hub

adithya-s-k /FineEnvs

FineEnvs — RL Environments 101: building and scaling RL environments in the age of LLMs

X-LANCE /Mobile-Env

A Universal Platform for Training and Evaluation of Mobile Interaction

andriusbern /stadium

A graphical interface for reinforcement learning and gym-based environments.

DeepGym /deepgym

RL environments with verifiable rewards you can trust. Sandboxed coding and computer-use tasks, verifiers kept apart from the answer, a QA gate against reward hacking, and infra failures kept out of training. Works with verl, TRL, OpenRLHF and Axolotl.

epignatelli /helx

Interoperating between (Deep) Reiforcement Learning libraries

Achronus /envrax

Gymnasium-style API standard for RL environment creation in JAX

allenai /pier

Workspace manager for coding agents. Interactively solve and develop Harbor tasks.

yunfanye /awesome-environment-scaling

Awesome Environment Scaling for AI Agents — a periodically updated survey and curated list of papers, projects, RL environments, world models, agent sandboxes, and open research problems

caiotheodoro /assay

An agentic auditor for RL environments and eval suites. Emits an Environment Card: every claim tied to a probe result, and every probe that could not run, with the reason.

aig-upf /gridenvs

Create new gridworld gym environments easily

nadeauglenn1-max /crucible

Turn any real software into a replayable RL environment for training AI agents — deterministic replay, verifiable rewards, TRL & verifiers adapters.

gumlau /awesome-workplace-agent-environments

A curated list of environments, benchmarks, datasets, and RL training resources for workplace AI agents: email, chat, calendar, docs, and enterprise apps.

prabhu-gopal /bohrin

Check the checker: an open standard for testing the graders of RL environments, with proof before every accusation.

Navitallabs /navital-gym

Deterministic, auto-graded environments for AI agents on Indian-reality tasks: UPI, GST, trip planning, forms, Hindi/Hinglish support. Rule-based grading, no LLM judge.

alt-glitch /sitor

Simulating Backend APIs for RL environments

carlostoledo1891 /cert-machine

Independent exact certification of machine-generated mathematics — exact arithmetic, no code shared with the claimant, refusal as a verdict.

meagoodboy /envdoctor

Adversarial QA for LLM-RL environments: find out what reward an empty answer earns. Model-free, zero API cost.