Research map / Image generation / Fast diffusion sampling
Rethinking Visual Rearrangement from A Diffusion Perspective
Tianliang Qi, Xinhang Song, Yuyi Liu, Shuqiang 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 |
|---|---|---|---|
| use diffusion-bridge denoising transformer instead of Kuhn-Munkres matching (ablation) | Diffusion Rearrangement · RoomR test | Success (%): 12.8 -> 17.8 | Improved |
| use denoising transformer instead of direct matching (ablation) | Diffusion Rearrangement · RoomR test | Success (%): 6.5 -> 17.8 | Improved |
| represent objects as Gaussian mixture distributions instead of center coordinates (ablation) | Diffusion Rearrangement · RoomR test | Success (%): 16.7 -> 17.8 | Improved |
| represent objects as Gaussian mixture distributions instead of PointNet++ features (ablation) | Diffusion Rearrangement · RoomR test | Success (%): 17.2 -> 17.8 | Improved |
| represent objects as GMM instead of coordinates (ablation) | Diffusion Rearrangement · RoomR test | Fixed Strict (%): 37.1 -> 38.8 | Improved |
| increase sampling steps from 20 to 200 (ablation) | Diffusion Rearrangement · RoomR validation | Success (%): 16.4 -> 17.9 | Improved |
| increase sampling steps beyond 200 (ablation) | Diffusion Rearrangement · RoomR validation | Success (%): 17.9 -> 17.4 (1000 steps) | Hurt |
| add Gaussian noise to depth input (ablation) | Diffusion Rearrangement · RoomR test | Success (%): 17.8 -> 17.5 | Hurt |
| diffusion-based inference instead of matching-based change inference | Diffusion Rearrangement vs CAVR · RoomR test | Success (%): 14.2 -> 17.8 | Improved |
| diffusion-based inference instead of matching-based change inference | Diffusion Rearrangement vs CAVR · ProcTHOR (self-built) test | Success (%): 4.9 -> 8.4 | Improved |
Closest papers by meaning
- Tracing the Roots: Leveraging Temporal Dynamics in Diffusion Trajectories for Origin Attribution (NeurIPS 2025 · similarity 0.88)
- Move Anything with Layered Scene Diffusion (CVPR 2024 · similarity 0.88)
- Diffusing Differentiable Representations (NeurIPS 2024 · similarity 0.87)
- The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise (ICLR 2025 · similarity 0.87)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.87)
- Temporal Score Analysis for Understanding and Correcting Diffusion Artifacts (CVPR 2025 · similarity 0.87)
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.
