Research map / Vision-language and multimodal / Robot action and navigation
ActiveVLA: Injecting Active Perception into Vision-Language-Action Models for Precise 3D Robotic Manipulation
Zhenyang Liu, Yongchong Gu, Yikai Wang, Xiangyang Xue, Yanwei Fu
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 active view selection (A-VS) (ablation) | fixed-view ActiveVLA · RLBench | success rate (%): 87.6 -> 89.4 | Improved |
| add active 3D zoom-in (A-3Z) on top of A-VS (ablation) | ActiveVLA · RLBench | success rate (%): 89.4 -> 91.8 | Improved |
| add active view selection (A-VS) (ablation) | fixed-view ActiveVLA · COLOSSEUM | success rate (%): 63.6 -> 64.5 | Improved |
| add active 3D zoom-in (A-3Z) on top of A-VS (ablation) | ActiveVLA · COLOSSEUM | success rate (%): 64.5 -> 65.9 | Improved |
| add active view selection (A-VS) (ablation) | fixed-view ActiveVLA · GemBench | success rate (%): 48.9 -> 49.4 | Improved |
| add active 3D zoom-in (A-3Z) on top of A-VS (ablation) | ActiveVLA · GemBench | success rate (%): 49.4 -> 51.3 | Improved |
| add active view selection (A-VS) inference cost (ablation) | fixed-view ActiveVLA · RLBench | inference time (s): 0.26 -> 0.45 | Hurt |
| add A-3Z inference cost (ablation) | ActiveVLA with A-VS · RLBench | inference time (s): 0.45 -> 0.53 | Hurt |
| increase number of views from one to three (ablation) | ActiveVLA · RLBench | success rate (%): 82.2 -> 91.8 | Improved |
| increase number of views beyond three (ablation) | ActiveVLA · RLBench | success rate (%) | No clear effect |
| increase zoom-in factor from 1 to 4 (ablation) | ActiveVLA · RLBench | success rate (%) | Improved |
| excessive zoom-in factor (ablation) | ActiveVLA · RLBench | success rate (%) | Hurt |
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.
- BridgeVLA — BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models (NeurIPS 2025) · better on 5 of 5 shared results
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
- BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models (NeurIPS 2025 · similarity 0.92)
- Localizing, Structuring, and Rendering: Bridging 3D and 2D Vision-Language-Action Models for Robotic Manipulation (CVPR 2026 · similarity 0.91)
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · similarity 0.91)
- ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation (NeurIPS 2025 · similarity 0.91)
- DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge (NeurIPS 2025 · similarity 0.90)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
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.
