Research map / Vision-language and multimodal / Robot action and navigation
Towards Efficient and Robust Manipulation via Multi-Frame Vision-Language-Action Modeling
Hao Li, Shuai Yang, Yilun Chen, Xinyi Chen, Xiaoda Yang, Yang Tian, Hanqing Wang, Tai Wang, Dahua Lin, Feng Zhao, Jiangmiao Pang
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 |
|---|---|---|---|
| feed multiple frames directly into the VLM backbone with discrete action tokens (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 31.0 -> 32.4 (WidowX VM 10.4 -> 4.2; speed 5.18 -> 3.09 Hz) | Mixed |
| replace discrete action tokens with continuous feature prediction and train decoder (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 32.4 -> 48.2; 3.09 -> 8.73 Hz | Improved |
| also train the vision-language backbone in multi-frame post-training (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 48.2 -> 67.2 | Improved |
| add multi-frame regularization (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 67.2 -> 70.9 | Improved |
| add cross-attention in decoder instead of self-attention (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 68.3 -> 70.9 | Improved |
| add feature modulator (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 63.5 -> 70.9 | Improved |
| use DiT decoder instead of MLP decoder (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 51.8 -> 70.9 | Improved |
| use diffusion objective (DiT) instead of flow matching (SiT) (ablation) | CronusVLA 7B · SimplerEnv avg | success rate (%): 67.6 -> 70.9 | Improved |
| increase number of input frames beyond optimum (ablation) | CronusVLA · SimplerEnv avg | success rate (%) | Hurt |
| multi-frame modeling with CronusVLA vs single-frame models | CronusVLA 7B · SimplerEnv-OR total avg | R-Score: 72.1 -> 86.9 (vs CogACT) | Improved |
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.
- TraceVLA — TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies (ICLR 2025) · better on 1 of 1 shared result
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
- Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.91)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge (NeurIPS 2025 · similarity 0.91)
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · similarity 0.91)
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.
