Wjjjh/lingbot-va-libero-goal

LingBot-VA LIBERO-Goal

This repository contains a community post-trained transformer checkpoint for LingBot-VA on the LIBERO-Goal suite. It is not an official Robbyant model release.

The checkpoint predicts robot actions jointly with future visual observations. It is intended for research and evaluation in the LIBERO simulator.

Checkpoint and result

Each run covers all 10 LIBERO-Goal tasks with 50 episodes per task. The first run used the runtime-default seed, which was not explicitly recorded; runs 2 and 3 used explicit seeds.

The population standard deviation across the three run-level success rates is 0.52 percentage points.

For reference, the official LingBot-VA repository reports 97.2% ± 0.2% for LIBERO-Goal. This community checkpoint's three-run mean is 0.47 percentage points lower. The two figures should not be treated as strictly identical experiments unless every training and evaluation detail is matched.

Model contents

This repository provides the post-trained transformer only. The VAE, tokenizer, and text encoder must be taken from `robbyant/lingbot-va-base`.

The checkpoint config is saved with:

{"attn_mode": "flex"}

flex is the training setting. For inference or evaluation, copy the checkpoint into a runtime model directory and change attn_mode to torch or flashattn, following the official LingBot-VA instructions. Do not overwrite the archived training checkpoint merely to switch inference backends.

Download

hf download Wjjjh/lingbot-va-libero-goal \
  --include "transformer/*" \
  --local-dir lingbot-va-libero-goal-step5000

Training details

Training followed the LingBot-VA LIBERO post-training pipeline, with a suite-specific action normalization computed from all LIBERO-Goal training actions.

Action and evaluation configuration

The model uses the first seven channels of the 30-channel action layout:

used_action_channel_ids = [0, 1, 2, 3, 4, 5, 6]
action_snr_shift = 0.05

The action quantiles used for both training and evaluation are:

{
  "q01": [
    -0.8973214030265808,
    -0.7473214268684387,
    -0.9375,
    -0.15214285254478455,
    -0.16256785675883295,
    -0.14142857491970062,
    -1.0
  ],
  "q99": [
    0.9375,
    0.9133928418159485,
    0.9375,
    0.21214285492897034,
    0.2582142949104309,
    0.375,
    1.0
  ]
}

The reported evaluations used:

• —LIBERO-Goal, 10 tasks
• —50 episodes per task (500 total)
• —maximum 800 environment steps per episode
• —128 × 128 agent-view and eye-in-hand observations
• —4 parallel GPU shards
• —one runtime-default seed and explicit seeds 2026072402 and 2026072403
• —LIBERO commit 8f1084e3132a39270c3a13ebe37270a43ece2a01

Limitations

• —The first of the three evaluation runs used an unrecorded runtime-default seed; only runs 2 and 3 are exactly seed-reproducible from this card.
• —Results are simulator- and configuration-dependent.
• —This model has not been validated for safe real-world robot deployment.
• —These files contain model weights only and do not contain optimizer state, so they are not a complete resumable training checkpoint.

Acknowledgements and citation

This checkpoint is derived from the official LingBot-VA base model and training code. Please cite the original LingBot-VA work when using this model:

@article{lingbot-va2026,
  title={Causal World Modeling for Robot Control},
  author={Li, Lin and Zhang, Qihang and Luo, Yiming and Yang, Shuai and Wang, Ruilin and Han, Fei and Yu, Mingrui and Gao, Zelin and Xue, Nan and Zhu, Xing and Shen, Yujun and Xu, Yinghao},
  journal={arXiv preprint arXiv:2601.21998},
  year={2026}
}