Research map / Vision-language and multimodal / Robot action and navigation
RA-VLA: Retrieval-Augmented VLA for Test-Time Adaptation
Sanghwan Jang, Minjin Jeon, Minsoo Kim, Seong Jin Choi, Dongha Kim, Hwanjo Yu
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 contextual adherence loss (ablation) | RA-VLA · LIBERO-Goal | success rate: 0.098 -> 0.532 | Improved |
| add contextual adherence loss (ablation) | RA-VLA · LIBERO-Goal | S_ctx: 0.0353 -> 0.3639 | Improved |
| replace off-the-shelf retriever with action-aware retriever (ablation) | RA-VLA · LIBERO-Goal | success rate: 10.2% -> 53.2% | Improved |
| use action-aware retriever instead of SigLIP 2 (ablation) | RAEA · LIBERO average | success rate: 0.0450 -> 0.1235 | Improved |
| use action-aware retriever instead of SigLIP 2 (ablation) | RICL_G · LIBERO average | success rate: 0.0885 -> 0.1295 | Improved |
| use action-aware retriever instead of SigLIP 2 (ablation) | RICL_R · LIBERO average | success rate: 0.0465 -> 0.2085 | Improved |
| RA-VLA instead of best baseline RICL_R | LIBERO unseen suite · LIBERO average | success rate: 20.9% -> 38.5% | Improved |
| RA-VLA instead of best baseline RICL_R | UR5e real robot · UR5e average | success rate: 35.4% -> 56.3% | Improved |
| increase buffer size N from 1 to 4 (ablation) | RA-VLA · LIBERO-Goal | success rate: 0.482 -> 0.552 | Improved |
| increase retrieved segments K from 1 to 4 (ablation) | RA-VLA · LIBERO-Goal | success rate: 0.532 -> 0.546 | No clear effect |
| decoupled per-segment encoding vs concatenating segments in prompt | RA-VLA · inference latency vs K | latency | Improved |
Closest papers by meaning
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model (ICML 2026 · similarity 0.92)
- VLA-Pro: Cross-Task Procedural Memory Transfer for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.91)
- 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.
