Research map / Vision-language and multimodal / Robot action and navigation
Transport Discrepancy as a Reliability Signal for Vision-Language-Action Models
Wanpeng Zhang, Ye Wang, Hao Luo, Haoqi Yuan, Yicheng Feng, Chaoyi Xu, Sipeng Zheng, Qin Jin, Zongqing Lu
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 |
|---|---|---|---|
| use sliced Wasserstein discrepancy instead of cosine distance (ablation) | pi0.5+DiG · LIBERO | avg success (%): 96.5 -> 98.3 | Improved |
| use sliced Wasserstein instead of MMD (ablation) | pi0.5+DiG · LIBERO | avg success (%): 96.8 -> 98.3 | Improved |
| use sliced Wasserstein instead of Sinkhorn (ablation) | pi0.5+DiG · LIBERO | avg success (%): 97.9 -> 98.3 | Improved |
| use discrepancy-driven gate instead of fixed gate g=0.5 (ablation) | pi0.5+DiG · LIBERO | avg success (%): 92.3 -> 98.3 | Improved |
| use discrepancy-driven gate instead of random gate (ablation) | pi0.5+DiG · LIBERO | avg success (%): 89.8 -> 98.3 | Improved |
| add residual refinement branch (lambda>0) to loss reweighting (ablation) | pi0.5+DiG · LIBERO | avg success (%): 97.3 -> 98.3 | Improved |
| use transport discrepancy gate instead of learned MLP gate (ablation) | pi0.5+DiG · LIBERO | avg success (%): 97.5 -> 98.3 | Improved |
| add DiG to pi0.5 | pi0.5 · LIBERO Long | success (%): 92.4 -> 96.4 | Improved |
| add DiG to GR00T-N1 | GR00T-N1 · LIBERO | avg success (%): 93.9 -> 95.3 | Improved |
| add DiG to pi0.5 in low-data regime | pi0.5 · RoboCasa 24 tasks (50 demos/task) | avg success (%): 41.4 -> 52.6 | Improved |
| add DiG under cosine+sine perturbation | pi0.5 · LIBERO perturbed | avg success (%): 82.2 -> 86.2 | Improved |
| add DiG on OOD Franka Spray-Plant | pi0.5 · Franka OOD Spray-Plant | whole-task success (%): 10 -> 30 | Improved |
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.
- CoT-VLA — CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025) · 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
- VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon (NeurIPS 2026 · similarity 0.90)
- EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models (NeurIPS 2025 · similarity 0.90)
- Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- CogVLA: Cognition-Aligned Vision-Language-Action Models via Instruction-Driven Routing & Sparsification (NeurIPS 2025 · similarity 0.90)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.88)
- Enabling VLA Action Self-Verification via VLM Token Probability Bucketing (NeurIPS 2026 · similarity 0.88)
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.
