Research map / Image generation / Fast diffusion sampling
SegDiff: Segmented Trajectory Diffusion for Consistent and Adaptive Robot Manipulation
Haidong Cao, Wenjun Cao, Quanhao Li, Sicheng Xie, Zhiying Du, Jiaqi Leng, Zuxuan Wu, Yu-Gang Jiang
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 |
|---|---|---|---|
| replace sliding-window DP with Segmented Trajectory Modeling | Diffusion Policy · RLBench (10 tasks) | avg success rate (%): 54.4 -> 86.7 | Improved |
| add Segmented Trajectory Modeling and DTE on full benchmark | Diffusion Policy / CoA · RLBench-60 | success rate: outperforming DP and CoA by 34% and 9% | Improved |
| add Dynamic Temporal Ensembling vs random sampling (ablation) | SegDiff · RLBench (10 tasks) | avg success rate: 0.880 -> 0.860 | No clear effect |
| add DTE vs random sampling (ablation) | SegDiff · Real-world Banana | success rate: 0.5 -> 0.9 | Improved |
| add DTE vs random sampling (ablation) | SegDiff · Real-world Package | success rate: 0.8 -> 1.0 | Improved |
| add DTE vs random sampling (ablation) | SegDiff · Real-world Screw Cap | success rate: 0.3 -> 0.6 | Improved |
| use INV only instead of random sampling (ablation) | SegDiff · RLBench (10 tasks) | avg success rate: 0.880 -> 0.800 | Hurt |
| use update factor eta=0 instead of 0.25 (ablation) | SegDiff · RLBench (10 tasks) | avg success rate: 0.860 -> 0.680 | Hurt |
| use update factor eta=1 instead of 0.25 (ablation) | SegDiff · RLBench (10 tasks) | avg success rate: 0.860 -> 0.800 | Hurt |
| reduce sampling and inversion steps to 10/10 (ablation) | SegDiff · RLBench (10 tasks) | success rate / time: 0.860 -> 0.764, 595.74 -> 121.95 ms | Mixed |
| reduce inversion steps from 50 to 10 (ablation) | SegDiff · RLBench (10 tasks) | success rate / time: 0.860 -> 0.860, 595.74 -> 380.21 ms | No clear effect |
| change DTE threshold from max to p99 (ablation) | SegDiff · RLBench (10 tasks) | avg success rate: 0.860 -> 0.836 | No clear effect |
Later papers that compared with this method
2 later accepted papers report SegDiff as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- RDF-MIG: A Robust Diffusion Framework for Masked Image Generation to Augment Semantic Segmentation and Change Detection (CVPR 2026 · better on 1 of 1 shared result)
- VGD: Value-Guided Diffusion Toward High-Utility Medical Image Segmentation (AAAI 2026 · 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
- ET-SEED: EFFICIENT TRAJECTORY-LEVEL SE(3) EQUIVARIANT DIFFUSION POLICY (ICLR 2025 · similarity 0.90)
- SkiP: When to Skip and When to Refine for Efficient Robot Manipulation (NeurIPS 2026 · similarity 0.90)
- Implicit Drifting Policy: One-Step Action Generation via Conditional Expert Geometry (NeurIPS 2026 · similarity 0.89)
- Latent Diffusion Planning for Imitation Learning (ICML 2025 · similarity 0.89)
- VidMan: Exploiting Implicit Dynamics from Video Diffusion Model for Effective Robot Manipulation (NeurIPS 2024 · similarity 0.88)
- Chain-of-Action: Trajectory Autoregressive Modeling for Robotic Manipulation (NeurIPS 2025 · 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.
