Research map / Vision-language and multimodal / Robot action and navigation
From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model
Bing Hu, Zaijing Li, Rui Shao, Junda Chen, April Hua Liu, Wei-Shi Zheng, Liqiang Nie
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 VBE and PBD on top of pi0.5 (ablation) | pi0.5 · Real-world Generalization | success rate: 57.0 -> 70.0 | Improved |
| add VBE and PBD on top of pi0.5 (ablation) | pi0.5 · Real-world Long-horizon | success rate: 41.0 -> 55.0 | Improved |
| add VBE and PBD on top of pi0.5 (ablation) | pi0.5 · LIBERO Long | success rate: 92.4 -> 94.6 | Improved |
| add VBE and PBD on top of pi0.5 (ablation) | pi0.5 · LIBERO Spatial | success rate: 98.8 -> 99.2 | Improved |
| add VBE alone or PBD alone (ablation) | pi0.5 · LIBERO Spatial | success rate: 98.8 -> 98.2 (one single-module row) | No clear effect |
| add VBE and PBD to pi0.5 on RoboTwin 2.0 Hard | pi0.5 · RoboTwin 2.0 Hard (20 tasks avg) | success rate: 49.2 -> 58.0 | Improved |
| add VBE and PBD to pi0.5 on LIBERO | pi0.5 · LIBERO average | success rate: 96.9 -> 98.0 | Improved |
| tune prior guidance strength lambda away from its optimum (ablation) | BehaviorVLA · LIBERO suites | success rate | Hurt |
| retrieve k=5 prototypes instead of 1, 3 or 10 (ablation) | BehaviorVLA · LIBERO Long | success rate: k=1 93.0, k=3 93.8, k=5 94.6, k=10 94.0 | Mixed |
| remove vision stream or action stream from VBE (ablation) | VBE · t-SNE of 20 task prototypes | cluster separability (qualitative) | Hurt |
| train with only 50% of real-world demonstrations | BehaviorVLA · Real-world tasks | success rate | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MemoryVLA — MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulation (ICLR 2026) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.93)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- Beyond Data Scaling: Representation-Centric Pre-training for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.91)
- RotVLA: Rotational Latent Action for Vision-Language-Action Model (NeurIPS 2026 · similarity 0.91)
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.
