Research map / Vision-language and multimodal / Robot action and navigation
RoboOmni: Actions Are Just Another Modality for Vision-Language Models
Dong Wang, Zilong Chen, Jirong Liu, Ziqing Qiao, Xin Xiao, Bingyi Kang, Hongtao Wu, Xiao Ma, Tao Kong, Huaping Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Multi-Token Action Prediction (MTAP) to FAST tokenizer (ablation) | RoboOmni(FAST) · CALVIN ABCD->D | 5-task success rate: 80.1% -> 88.1% | Improved |
| add MTAP to Bin tokenizer (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | 5-task success rate: 67.9% -> 83.4% | Improved |
| add MTAP parallel decoding to Bin tokenizer for speed (ablation) | RoboOmni(Bin) · CALVIN | inference time (ms/action): 107 -> 12.1 | Improved |
| increase history window from 1 to 5 (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | 5-task success rate: 81.3% -> 83.4% | Improved |
| increase history window from 5 to 10 (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | average task length: 4.64 -> 4.55 | No clear effect |
| scale backbone from 2B to 7B (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | 5-task success rate: 77.6% -> 83.4% | Improved |
| add VLM data co-training (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | average task length: 4.53 -> 4.64 | Improved |
| add sequence packing (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | average task length: 4.46 -> 4.64 | Improved |
| add classifier-free guidance training (drop language 20%) (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | average task length: 4.48 -> 4.64 | Improved |
| use 1024 action bins instead of 256 (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | 5-task success rate: 83.4% -> 79.0% | Hurt |
| use 128 action bins instead of 256 (ablation) | RoboOmni(FAST) · CALVIN ABCD->D | 5-task success rate: 88.1% -> 86.1% | Hurt |
| use 128 bins instead of 256 for Bin tokenizer (ablation) | RoboOmni(Bin) · CALVIN ABCD->D | 5-task success rate: 83.4% -> 83.7% | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UP-VLA — UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent (ICML 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FASTer: Toward Powerful and Efficient Autoregressive Vision–Language–Action Models with Learnable Action Tokenizer and Block-wise Decoding (ICLR 2026 · similarity 0.93)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.90)
- Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
