Research map / Vision-language and multimodal / Robot action and navigation
Tabero: Learning Gentle Manipulation with Closed-Loop Force Feedback from Vision, Touch, and Language
Qiwei Wu, Rui Zhang, Xin Xiang, Tao Li, Weihua Zhang, Junjie Lai, Renjing Xu
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 (force field, image or force MLP) to the VLA policy (ablation) | Tabero-VTLA (Pi0 + LoRA) · Tabero object subset, gentle instruction | success rate: 0.00 -> 0.01 (gentle) and 0.00 -> 0.40 (firm) for Field | Improved |
| add force supervision loss to the field-based tactile model (ablation) | Tabero-VTLA · Tabero object subset, gentle instruction | success rate: 0.01 -> 0.52 | Improved |
| add tactile marker field on top of force supervision (ablation) | Tabero-VTLA · Tabero object subset, gentle instruction | success rate: 0.45 -> 0.52 | Improved |
| use marker field instead of tactile image as tactile representation (with force supervision) (ablation) | Tabero-VTLA · Tabero object subset, gentle instruction | success rate: 0.48 -> 0.52 | No clear effect |
| inject force via decoder (TA-VLA style) instead of marker field with force supervision (ablation) | Tabero-VTLA · Tabero object subset, gentle instruction | success rate: 0.52 -> 0.31 | Hurt |
| remove feedforward term from hybrid grip controller (ablation) | hybrid grip force controller · Tabero Object task 1, 25% force | grip force tracking | Hurt |
| remove admittance component from hybrid grip controller (ablation) | hybrid grip force controller · Tabero Object task 1, 25% force | grip force tracking | Hurt |
| replace tactile-enabled continuous control with vision-language only on binary gripper data (ablation) | Tabero-VTLA · Tabero object subset, gentle instruction (10% force) | average grip force (N): 3.7 -> 34.6 | Hurt |
| reduce gentle force target from 25% to 10% (ablation) | Tabero-VTLA · Tabero object subset, gentle instruction | success rate: 0.79 -> 0.52 | Hurt |
| use tactile gripper with reduced force settings during data replay (ablation) | data collection in Isaac Lab · LIBERO four-task average | success rate: 0.85 -> 0.36 | Hurt |
| test unseen adverbs lightly and forcefully | Tabero-VTLA · Tabero object subset (Dataset B) | average grip force (N): lightly 14.5, forcefully 18.1 | Mixed |
Closest papers by meaning
- ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation (NeurIPS 2025 · similarity 0.89)
- Touch in the Wild: Learning Fine-Grained Manipulation with a Portable Visuo-Tactile Gripper (NeurIPS 2025 · similarity 0.88)
- Object-Centric Prompt-Driven Vision-Language-Action Model for Robotic Manipulation (CVPR 2025 · similarity 0.86)
- Enhancing Tactile-based Reinforcement Learning for Robotic Control (NeurIPS 2025 · similarity 0.86)
- ForceFlow: Learning to Feel and Act via Contact-Driven Flow Matching (NeurIPS 2026 · similarity 0.86)
- Touch-R1: Reinforcing Touch Reasoning in MLLMs (NeurIPS 2026 · similarity 0.86)
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.
