Research map / Vision-language and multimodal / Robot action and navigation
SCALE: Self-uncertainty Conditioned Adaptive Looking and Execution for Vision-Language-Action Models
Hyeonbeom Choi, Daechul Ahn, Youhan Lee, Taewook Kang, Seongwon Cho, Jonghyun Choi
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 adaptive action decoding (ablation) | SCALE · LIBERO-Long (OpenVLA) | success rate (%): 52.7 -> 58.0 | Improved |
| add adaptive visual attention (ablation) | SCALE · LIBERO-Long (OpenVLA) | success rate (%): 52.7 -> 56.0 | Improved |
| combine adaptive decoding and adaptive visual attention (ablation) | SCALE · LIBERO-Long (OpenVLA) | success rate (%): 52.7 -> 63.3 | Improved |
| use dual-reference self-uncertainty instead of Self-Certainty (ablation) | SCALE · LIBERO-Long (OpenVLA) | success rate (%): 53.8 -> 63.3 | Improved |
| use self-uncertainty instead of entropy (ablation) | SCALE · LIBERO-Long (OpenVLA) | success rate (%): 55.4 -> 63.3 | Improved |
| modulate vision encoder attention instead of VLA cross-modal attention (ablation) | SCALE · LIBERO-Long (OpenVLA) | success rate (%): 57.4 -> 63.3 | Improved |
| use continuous adaptive scaling instead of binary fixed switch (ablation) | SCALE visual modulation · LIBERO-Long (OpenVLA) | success rate (%): 54.8 -> 57.4 | Improved |
| use uncertainty change from EMA instead of instantaneous uncertainty (ablation) | SCALE visual modulation · LIBERO-Long (OpenVLA) | success rate (%): 55.4 -> 63.3 | Improved |
| apply fixed-hyperparameter sampling (temperature, top-k, top-p) instead of greedy | pi0-FAST · LIBERO (Avg.) | success rate (%): 91.2 -> 84.3/88.1/86.2 | Hurt |
| apply fixed-hyperparameter sampling instead of greedy | OpenVLA · LIBERO (Avg.) | success rate (%): 75.7 -> 76.7/76.7/77.2 | Improved |
| add SCALE to OpenVLA | OpenVLA · LIBERO (Avg.) | success rate (%): 75.7 -> 81.5 | Improved |
| add SCALE to SpatialVLA | SpatialVLA (zero-shot) · SIMPLER-WidowX (Avg.) | success rate (%): 31.3 -> 41.7 | Improved |
Closest papers by meaning
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- Beyond Data Scaling: Representation-Centric Pre-training for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.90)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.89)
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model (ICML 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.
