Research map / Vision-language and multimodal / Robot action and navigation
ForceVLA2: Unleashing Hybrid Force-Position Control with Force Awareness for Contact-Rich Manipulation
Yang Li, Zhaxizhuoma Zhaxizhuoma, Hongru Jiang, Junjie Xia, Hongquan Zhang, Jinda Du, Yunsong Zhou, Jia Zeng, Ce Hao, Jieji Ren, Qiaojun Yu, Cewu Lu and 2 more
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 force prompt to π0 (ablation) | π0 · ForceVLA2-Dataset avg | success rate (%): 18.0 -> 27.0 | Improved |
| add multimodal encoder on top of force prompt (ablation) | π0 + force prompt · ForceVLA2-Dataset avg | success rate (%): 27.0 -> 40.0 | Improved |
| add Cross-Scale MoE (ablation) | π0 + force prompt + multimodal encoder · ForceVLA2-Dataset avg | success rate (%): 40.0 -> 66.0 | Improved |
| add force modality input to Cross-Scale MoE (ablation) | ForceVLA2 · ForceVLA2-Dataset avg | success rate (%): 50.0 -> 66.0 | Improved |
| add visual modality input to Cross-Scale MoE (ablation) | ForceVLA2 · ForceVLA2-Dataset avg | success rate (%): 36.0 -> 50.0 | Improved |
| add extra force token at MoE input (ablation) | ForceVLA2 · Press bottle / Clean board | success rate (%): 85.0 -> 80.0 (Press bottle), 85.0 -> 70.0 (Clean board) | Hurt |
| naively concatenate force input to π0 | π0 · ForceVLA2-Dataset avg | success rate (%): 18.0 -> 17.0 | No clear effect |
| use ForceVLA2 instead of ForceVLA | contact-rich VLA · ForceVLA2-Dataset avg | success rate (%): 35.0 -> 66.0 | Improved |
| use ForceVLA2 instead of π0.5 | contact-rich VLA · ForceVLA2-Dataset avg | success rate (%): 31.0 -> 66.0 | 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.
- ForceVLA — ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation (NeurIPS 2025) · better on 6 of 6 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
- ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation (NeurIPS 2025 · similarity 0.96)
- ForceFlow: Learning to Feel and Act via Contact-Driven Flow Matching (NeurIPS 2026 · similarity 0.90)
- Tabero: Learning Gentle Manipulation with Closed-Loop Force Feedback from Vision, Touch, and Language (ICML 2026 · similarity 0.89)
- Object-Centric Prompt-Driven Vision-Language-Action Model for Robotic Manipulation (CVPR 2025 · similarity 0.88)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.88)
- PAC Bench: Do Foundation Models Understand Prerequisites for Executing Manipulation Policies? (NeurIPS 2025 · similarity 0.88)
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.
