InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization

InternVLA-A1.5 unifies vision-language understanding, latent visual foresight, and action generation in one robot policy. It builds on a native Qwen3.5-2B VLM backbone, preserves semantic learning through VQA and subtask prediction, and attaches a lightweight unified action expert for continuous control.

This repository hosts InternVLA-A1.5-DOMINO, the checkpoint fine-tuned for DOMINO dynamic manipulation evaluation. It corresponds to the DOMINO SFT result reported for InternVLA-A1.5, where the model is adapted on the DOMINO ALOHA-AgileX Level-1 training split and evaluated on the clean Level-1 DOMINO suites.

Covering base and benchmark-specific checkpoints, we release the InternVLA-A1.5 series:

InternVLA-A1.5-base: base checkpoint for downstream fine-tuning and evaluation
InternVLA-A1.5-RoboTwin: fine-tuned on RoboTwin 2.0
InternVLA-A1.5-Libero: fine-tuned on LIBERO
InternVLA-A1.5-DOMINO: fine-tuned on DOMINO

🔑 Key Features

🔮 The Core: Attaches a lightweight unified action expert to a native Qwen3.5-2B VLM backbone through shared full-attention layers, while preserving modality-specific Gated DeltaNet processing.
🚀 The Foresight: Uses learnable foresight tokens to query task-relevant future dynamics from the shared multimodal context, supervised by a frozen WAN2.2-5B video generation model during training.
⚡ The Output: Discards the video branch at inference and predicts continuous action chunks through flow matching, keeping deployment latency practical.

Model Details

Model type: Vision-Language-Action robot policy
Base checkpoints: InternRobotics/InternVLA-A1.5-base and InternRobotics/InternVLA-A1.5-RoboTwin
Backbone: Qwen/Qwen3.5-2B
Policy type: internvla_a1_5
Fine-tuning target: DOMINO ALOHA-AgileX Level-1 training split
Evaluation scope: DOMINO SFT / dynamic-to-dynamic evaluation
Action head: unified action expert with flow-matching action generation
State/action dimension: up to 32
Image resolution: 224 x 224
License: CC BY-NC-SA 4.0

Usage

Please refer to our official repo InternVLA-A-series for installation, training, fine-tuning, and evaluation.

For DOMINO fine-tuned evaluation:

git clone https://github.com/InternRobotics/InternVLA-A-series.git
cd InternVLA-A-series
bash evaluation/DOMINO/eval.sh \
  InternRobotics/InternVLA-A1.5-DOMINO \
  outputs/domino/internvla_a1_5_domino_sft \
  demo_clean_dynamic \
  8 \
  fm \
  50 \
  100 \
  100000 \
  10 \
  abs \
  float32 \
  10

For benchmark workflows, please see:

Demonstrations

InternVLA-A1.5-DOMINO corresponds to the DOMINO SFT result. SR is the primary success-rate metric, and MS denotes manipulation score.

DOMINO Setting	SR (%) ↑	MS ↑
Fine-tuned / SFT dynamic manipulation	29.3	42.5

For reference, the RoboTwin-tuned checkpoint reaches 27.7 SR and 39.8 MS in the DOMINO zero-shot setting.

License and Citation

All code within this repo is released under CC BY-NC-SA 4.0. Please consider citing our project if it helps your research.

@article{internvla_a15,
  title={InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization},
  author={Ma, Haoxiang and Cai, Junhao and Xu, Xiaoxu and Li, Hao and Yang, Yuyin and Tian, Yang and Cao, Jiafei and Zhu, Hongrui and Qiu, Zherui and Zhaxizhuoma and Yang, Yuqiang and Peng, Jiaqi and Wei, Xueyuan and Zhu, Yangkun and Jiang, Jiahao and Gao, Xing and Wang, Hanqing and Yuan, Feng and Li, Kailin and Zhu, Xueyue and Wang, Tai and Ding, Yan and Pang, Jiangmiao and Zeng, Jia and Zhang, Jingjing and Zhou, Bowen and Mu, Yao and Shen, Chunhua and Zhang, Weinan},
  journal={arXiv preprint arXiv:2607.04988},
  year={2026}
}