Research map / Vision-language and multimodal / Robot action and navigation
Restoring Linguistic Grounding in VLA Models via Train-Free Attention Recalibration
Ninghao Zhang, Bin Zhu, Shijie Zhou, Jingjing Chen
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 IGAR attention recalibration | pi0 · LIBERO-Goal ICBench V4 | SR under contradiction (%): 76.4 -> 36.4 | Improved |
| add IGAR attention recalibration | OpenVLA-OFT · LIBERO-Goal ICBench V1 | SR under contradiction (%): 97.8 -> 66.4 | Improved |
| add IGAR attention recalibration | pi0.5 · LIBERO ICBench Spatial | SR under contradiction (%): 96.2 -> 95.8 | No clear effect |
| add IGAR attention recalibration | pi0 · LIBERO normal instructions | SR (%): 97.1 -> 96.7 | No clear effect |
| add IGAR attention recalibration | OpenVLA-OFT · LIBERO normal instructions | SR (%): 98.0 -> 98.5 | No clear effect |
| add IGAR attention recalibration | pi0.5 · LIBERO-Object normal instructions | SR (%): 98.4 -> 94.4 | Hurt |
| add attention redistribution after sink detection and head selection (ablation) | pi0.5 · LIBERO-Goal ICBench V1 | LGS: 3.8 -> 7.4 | Improved |
| add sink detection to head selection and redistribution (ablation) | pi0.5 · LIBERO-Goal ICBench V1 | LGS: 4.6 -> 7.4 | Improved |
| add grounding head selection (ablation) | pi0.5 · LIBERO-Goal ICBench V1 | LGS: 4.2 -> 7.4 | Improved |
| set text-sink decay p outside 0.3-0.8 (ablation) | IGAR on OpenVLA-OFT · LIBERO-Goal | LGS | Hurt |
| intervene on more than 24 layers (ablation) | IGAR on OpenVLA-OFT · LIBERO-Goal | LGS | Hurt |
| IGAR inference overhead | OpenVLA-OFT · per-step latency | ms: 98.7 -> 105.6 | Hurt |
Closest papers by meaning
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.93)
- Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.93)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.92)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.92)
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon (NeurIPS 2026 · similarity 0.92)
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.
