# alibaba-damo-academy/RynnVLA-002

RynnVLA-002: A Unified Vision-Language-Action and World Model

Repository: https://github.com/alibaba-damo-academy/RynnVLA-002
Canonical: https://ross.abutalabs.com/products/rynnvla-002
Language: Python
License Family: other
Last push: 2025-12-02T10:34:06+00:00

## Health v2 (maintenance only)
Score: 43/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 55, release rhythm 35, longevity 31
- inputs: {"age_days": 436, "days_push": 274, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1119, forks 66 (observed 2026-08-28T04:03:39.350632+00:00)

## What it is
RynnVLA-002 is a unified autoregressive Vision-Language-Action and world model that generates robot actions from text and image observations and predicts future frames from actions. It includes model checkpoints, training code, and evaluation code for the LIBERO simulation benchmark and real-world LeRobot experiments.

## Use cases
- train a vision-language-action model for robot manipulation
- evaluate robot policies on the LIBERO benchmark
- generate next-frame predictions with a world model
- run real-world robot experiments with LeRobot
- unify action generation and image generation in one model

## When to choose
- you need a state-of-the-art VLA model for robotic manipulation research
- you want both action generation and world-model video prediction in a single framework
- you need LIBERO benchmark evaluation code and pretrained checkpoints

## When to avoid
- you need a production robot control system rather than a research codebase
- you lack GPU resources for large autoregressive model training
- you need a permissively licensed dependency - the license field is unclear despite an Apache 2.0 badge

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, computer-vision, robotics, llm-training
- domain: robotics, artificial-intelligence, deep-learning, computer-vision
- platform: python
- tags: vision-language-action, world-model, vla, robot-learning, autoregressive-model, libero-benchmark, lerobot, linux, gpu

## Member repositories
- alibaba-damo-academy/RynnVLA-002 (main) score 43

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:39.350632+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:41:09.980514+00:00, confidence not recorded.
  - readme: https://github.com/alibaba-damo-academy/RynnVLA-002 (fetched 2026-08-28T04:03:39.350632+00:00, sha ed9c14fca383)
- Data as of 2026-08-30T08:39:29.467469+00:00.