Research map / Vision-language and multimodal / Robot action and navigation
AT-VLA: Adaptive Tactile Injection for Enhanced Feedback Reaction in Vision-Language-Action Models
Xiaoqi Li, Muhe Cai, Jiadong Xu, Juan Zhu, Hongwei Fan, Yan Shen, Guangrui Ren, Hao Dong
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 tactile input directly without gating (ablation) | GO-1 · Contact-rich tasks average | Average success rate: 0.22 -> 0.13 | Hurt |
| add tactile gate with adaptive cross attention (ablation) | AT-VLA · Contact-rich tasks average | Average success rate: 0.13 -> 0.39 | Improved |
| add tactile reaction dual-stream (ablation) | AT-VLA · Contact-rich tasks average | Average success rate: 0.39 -> 0.50 | Improved |
| use force 6D instead of marker 2D tactile format (ablation) | AT-VLA · Contact-rich tasks average | Average success rate: 0.32 -> 0.50 | Improved |
| use force 6D instead of visual-tactile image format (ablation) | AT-VLA · Contact-rich tasks average | Average success rate: 0.40 -> 0.50 | Improved |
| add gate and dual-stream to marker 2D direct incorporation (ablation) | AT-VLA · Contact-rich tasks average | Average success rate: 0.05 -> 0.32 | Improved |
| add gate and dual-stream to V-T image direct incorporation (ablation) | AT-VLA · Contact-rich tasks average | Average success rate: 0.02 -> 0.40 | Improved |
| add adaptive tactile injection to GO-1 | GO-1 · Unzip Bag | Overall success rate: 0.20 -> 0.33 | Improved |
| add adaptive tactile injection to GO-1 | GO-1 · Wipe Vase | Overall success rate: 0.07 -> 0.67 | Improved |
| drop tactile input at inference for AT-VLA | AT-VLA · Stamp (modality-agnostic eval) | Success rate: 0.46 -> 0.20 | Hurt |
| train with tactile but infer without it | AT-VLA · Pick Place and Open Drawer | Success rate: 1.0 -> 1.0; 0.93 -> 0.93 | No clear effect |
Closest papers by meaning
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.90)
- ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation (NeurIPS 2025 · similarity 0.90)
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model (ICML 2026 · similarity 0.89)
- ATI-VLA: Action-Centric Predictive Vision–Language–Action Models via Actionable Alignment Then Adaptive Injection (NeurIPS 2026 · similarity 0.89)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.89)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · 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.
