Research map / Vision-language and multimodal / Robot action and navigation
Adaptive Action Chunking at Inference-time for Vision-Language-Action Models
Yuanchang Liang, Xiaobo Wang, Kai Wang, Shuo Wang, Xiaojiang Peng, Haoyu Chen, David Kim Huat Chua, Prahlad Vadakkepat
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 AAC adaptive chunking | GR00T N1.5 · RoboCasa (24 tasks) | success rate (%): 59.7 -> 62.0 | Improved |
| add AAC adaptive chunking | GR00T N1.5 · RoboCasa (Rotation) | success rate (%): 57.6 -> 61.4 | Improved |
| add AAC adaptive chunking | GR00T N1.5 · RoboCasa (Button) | success rate (%): 80.3 -> 81.3 | No clear effect |
| add AAC adaptive chunking | GR00T N1.5 · LIBERO (avg) | success rate (%): 94.1 -> 95.0 | Improved |
| add AAC adaptive chunking | GR00T N1.5 · LIBERO-Long | success rate (%): 88.8 -> 92.8 | Improved |
| add AAC adaptive chunking | pi0.5 · LIBERO (avg) | success rate (%): 97.0 -> 97.9 | Improved |
| add AAC adaptive chunking | pi0.5 · LIBERO-Pro (avg) | success rate (%): 30.9 -> 34.8 | Improved |
| add AAC adaptive chunking | GR00T N1.5 · LIBERO-Pro (avg) | success rate (%): 3.9 -> 6.3 | Improved |
| add AAC adaptive chunking | GR00T N1.5 · Real-world (3 tasks avg) | success rate (%): 67.0 -> 82.0 | Improved |
| use adaptive instead of fixed chunk size 16 (ablation) | GR00T N1.5 · RoboCasa (24 tasks) | success rate (%): 59.7 -> 62.0 | Improved |
| use fixed chunk size 8 instead of 16 (ablation) | GR00T N1.5 · RoboCasa (24 tasks) | success rate (%): 59.7 -> 61.2 | Improved |
| use fixed chunk size 2 instead of 16 (ablation) | GR00T N1.5 · RoboCasa (24 tasks) | success rate (%): 59.7 -> 47.0 | Hurt |
Closest papers by meaning
- Real-Time Execution of Action Chunking Flow Policies (NeurIPS 2025 · similarity 0.92)
- Bidirectional Decoding: Improving Action Chunking via Guided Test-Time Sampling (ICLR 2025 · similarity 0.90)
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model (ICML 2026 · similarity 0.89)
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
- VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon (NeurIPS 2026 · similarity 0.89)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.89)
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.
