Research map / Detection and segmentation / Open-set and few-shot detection
XSemanticFlow: Cross Object Semantic Alignment for Zero-shot Manipulation
Junyu Nan, Noam Eshed, Brian Okorn, Kris Kitani
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 rotation-invariant backbone fine-tuning (ablation) | XSemanticFlow · PartNet-Coseg | mIoU: 58.74 -> 63.32 | Improved |
| add rotation-invariant backbone fine-tuning (ablation) | XSemanticFlow · PartNet-SE3 | mIoU: 61.62 -> 62.47 | Improved |
| add supervised PartNet-Coseg fine-tuning (ablation) | XSemanticFlow · PartNet-Coseg | mIoU: 61.81 -> 63.32 | Improved |
| add geometric features to the cross-attention (ablation) | XSemanticFlow · PartNet-Coseg | mIoU: 63.32 -> 62.21 | Hurt |
| add geometric features to the cross-attention (ablation) | XSemanticFlow · PartNet-SE3 | Acc: 76.18 -> 75.10 | Hurt |
| replace PartField with XSemanticFlow aligned features | PartField · PartNet-SE3 | mIoU: 43.41 -> 62.47 | Improved |
| replace PartField with XSemanticFlow aligned features | PartField · PartNet-Coseg | mIoU: 52.43 -> 63.32 | Improved |
| use correspondence-based remapping instead of video-generation planning | NovaPlan · Mug Hanging | success rate: 20% -> 80% | Improved |
| use correspondence-based remapping instead of video-generation planning | NovaPlan · Block Pouring | success rate: 40% -> 70% | Improved |
| use correspondence-based remapping instead of NovaPlan | NovaPlan · Water Plant | success rate: 90% -> 90% | No clear effect |
Closest papers by meaning
- G3Flow: Generative 3D Semantic Flow for Pose-aware and Generalizable Object Manipulation (CVPR 2025 · similarity 0.88)
- FlowHOI: Flow-based Semantics-Grounded Generation of Hand-Object Interactions for Dexterous Robot Manipulation (NeurIPS 2026 · similarity 0.86)
- VidBot: Learning Generalizable 3D Actions from In-the-Wild 2D Human Videos for Zero-Shot Robotic Manipulation (CVPR 2025 · similarity 0.86)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.85)
- DenseMatcher: Learning 3D Semantic Correspondence for Category-Level Manipulation from a Single Demo (ICLR 2025 · similarity 0.85)
- SparseDFF: Sparse-View Feature Distillation for One-Shot Dexterous Manipulation (ICLR 2024 · similarity 0.85)
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.
