Research map / Vision-language and multimodal / Robot action and navigation
LAST: Bridging Vision-Language and Action Manifolds via Gromov-Wasserstein Alignment
Huaihai Lyu, Chaofan Chen, Yuheng Ji, Xiansheng Chen, Pengwei Wang, Shanghang Zhang, Changsheng Xu
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 se(3) Lie-algebra mapping (ablation) | LAST tokenizer · LIBERO Spatial | success rate (%): 88.5 -> 94.1 | Improved |
| add covariance-aware whitening (ablation) | LAST tokenizer · LIBERO Object | success rate (%): 93.5 -> 98.7 | Improved |
| add covariance-aware whitening (ablation) | LAST tokenizer · LIBERO Goal | success rate (%): 89.2 -> 94.6 | Improved |
| add commitment loss (lambda1=0 -> 0.5) (ablation) | LAST tokenizer · LIBERO Long | success rate (%): 66.2 -> 86.0 | Improved |
| set dominant commitment loss lambda1=10 instead of 0.5 (ablation) | LAST tokenizer · LIBERO Long | success rate (%): 86.0 -> 26.4 | Hurt |
| use DiT head instead of MLP head (ablation) | LAST-Continuous · Real-world ZipSeal | success rate (%): 52% -> 57% | Improved |
| use DiT head instead of MLP head (ablation) | LAST-Continuous · Real-world TubeRack | success rate (%): 43% -> 48% | Improved |
| use DiT head instead of MLP head (ablation) | LAST-Continuous · Real-world PlaceObj | success rate (%): 80% -> 73% | Hurt |
| add discrete supervision L_dis (ablation) | LAST VLA policy · Real-world ZipSeal | success rate (%): 22% -> 57% | Improved |
| add discrete supervision L_dis (ablation) | LAST VLA policy · Real-world TubeRack | success rate (%): 18% -> 48% | Improved |
| add discrete supervision L_dis (ablation) | LAST VLA policy · Real-world PlaceObj | success rate (%): 45% -> 73% | Improved |
| add continuous DiT refinement to FAST | pi0-FAST · LIBERO average | success rate (%): 85.5 -> 94.1 | Improved |
Closest papers by meaning
- 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)
- RotVLA: Rotational Latent Action for Vision-Language-Action Model (NeurIPS 2026 · similarity 0.90)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- ATI-VLA: Action-Centric Predictive Vision–Language–Action Models via Actionable Alignment Then Adaptive Injection (NeurIPS 2026 · 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.
