vlmrl
An experimental reinforcement learning gym for vision-language models, written in JAX. It lets you plug in environments and models and train them with PPO.
Share on XLicense: Apache-2.0
Overview
vlm-gym is an experimental reinforcement learning gym for vision-language models, written in JAX. The idea is to drop in any environment and any model and train with PPO. It ships with environments such as GeoGuessr, NLVR2 and captioning, and a reference Qwen3-VL-4B-Instruct model that can be converted from Hugging Face to JAX. The trainer, rollout engine and evaluation harness live in a core folder, and GeoGuessr training uses a staged curriculum.
Key features
- Pluggable vision environments: GeoGuessr, NLVR2, captioning
- PPO trainer for vision-language models
- Rollout engine and evaluation against a Hugging Face baseline
- Converts Qwen3-VL-4B-Instruct from Hugging Face to JAX
Best for
Suited to researchers experimenting with reinforcement learning on vision-language models in JAX. The README marks its status as experimental.
- Upstream
- sdan/vlm-gym
- Fork on GitHub
- Guo-astro/vlmrl
- Upstream stars
- 149
- Category
- AI agents and LLM tools
- Language
- Python
- License
- Apache-2.0
- Forked
- 2025-10-14
- Sync status
- In syncLast synced 2026-09-29
More in AI agents and LLM tools
A free desktop app where children in non-English-speaking countries practice English speaking and listening by talking with an AI companion. It can run locally and offline.
Forked 2026-09-28Last synced 2026-09-29License: MITAI agents and LLM toolsGitHub
An agent skill for Chinese patents. It helps find patentable points, write disclosure documents, explain patents in plain language, follow policy changes and assist with examination replies.
Forked 2026-09-28Last synced 2026-09-29License: MITAI agents and LLM toolsGitHub
A minimal Python library from Hugging Face for building AI agents that write and run code to solve tasks. It lets you set up capable agents in a few lines.
Forked 2026-09-28Last synced 2026-09-29License: Apache-2.0AI agents and LLM toolsGitHub
An open-source benchmark that measures how well AI agents handle realistic legal work. It provides tasks and evaluation for improving legal agents.
Forked 2026-09-28Last synced 2026-09-29License: MITAI agents and LLM toolsGitHub