Research map / Vision-language and multimodal / Robot action and navigation
CulinaryCut: A Physics-aware Vision-Language-Action Benchmark for Food Cutting via Material Point Method
Hyunsuh Koh, Changyong Song, Youngjae Choi, Misa Viveiros, David Hyde, Heewon Kim
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 |
|---|---|---|---|
| train on IntSim (MPM-filtered) trajectories instead of ManiSkill geometric trajectories (ablation) | RDT · orange cutting, IntSim eval | success rate (%): 23% -> 59% | Improved |
| train on IntSim trajectories instead of ManiSkill trajectories (ablation) | RDT · orange cutting, ManiSkill eval | success rate (%): 62% -> 68% | No clear effect |
| train on IntSim trajectories instead of ManiSkill trajectories (ablation) | RDT · orange cutting, real robot | successful cuts out of 10: 1/10 -> 3/10 | Improved |
| add more objects to the scene (multi-object) | RDT · CulinaryCut | success rate (%): 69% -> 31% | Hurt |
| add more objects to the scene (multi-object) | Octo · CulinaryCut | success rate (%): 42% -> 17% | Hurt |
| add more objects to the scene (multi-object) | OpenVLA · CulinaryCut | success rate (%): 51% -> 28% | Hurt |
| evaluate on an unseen object category (leave-one-object-out) | RDT · CulinaryCut | success rate (%): 69% -> 43% | Hurt |
| evaluate on an unseen object category | Octo · CulinaryCut | success rate (%): 42% -> 34% | Hurt |
| evaluate on an unseen object category | OpenVLA · CulinaryCut | success rate (%): 51% -> 26% | Hurt |
| flip cut direction at test (train 0.25 right, test 0.25 left) | RDT · CulinaryCut ratio cut | success rate (%): 55% -> 20% | Hurt |
| change ratio at test (train 0.75, test 0.25) | RDT · CulinaryCut ratio cut | success rate (%): 35% -> 0% | Hurt |
| change ratio at test (train 0.5, test 0.25) | RDT · CulinaryCut ratio cut | success rate (%): 60% -> 25% | Hurt |
Closest papers by meaning
- RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation Skills (NeurIPS 2025 · similarity 0.86)
- ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation (NeurIPS 2025 · similarity 0.85)
- Localizing, Structuring, and Rendering: Bridging 3D and 2D Vision-Language-Action Models for Robotic Manipulation (CVPR 2026 · similarity 0.85)
- DeformGen: Dynamics-Based Topology Augmentation for Deformable Manipulation Policy Learning (NeurIPS 2026 · similarity 0.84)
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · similarity 0.84)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.84)
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.
