Research map / 3D vision / 3D generation
Any2Any 3D Diffusion Models with Knowledge Transfer: A Radiotherapy Planning Study
Yuhan Wang, Zihan Li, Han Liu, Simon Arberet, Martin Kraus, Yuyin Zhou, Florin-Cristian Ghesu, Dorin Comaniciu, Ali Kamen, Riqiang Gao
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 pretrained Wan weights (ablation) | DiffKT3D DiT · GDP-HMM validation | MAE / Score: 2.58 / 119.82 -> 2.07 / 135.41 | Improved |
| switch from conditional to Any2Any training (ablation) | DiffKT3D DiT · GDP-HMM validation | MAE / Score: 2.07 / 135.41 -> 1.90 / 136.22 | Improved |
| add role embedding (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 2.01 / 134.04 -> 1.90 / 136.22 | Improved |
| add full attention (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 2.15 / 130.80 -> 1.90 / 136.22 | Improved |
| add modality-specific patch embedding (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 2.02 / 134.71 -> 1.90 / 136.22 | Improved |
| add 4D RoPE (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 1.96 / 135.27 -> 1.90 / 136.22 | Improved |
| add ScardNFT RL post-training (ablation) | Any2Any DiT · GDP-HMM validation | Scorecard: 136.22 -> 138.17 | Improved |
| add ScardNFT RL post-training (ablation) | Any2Any DiT · GDP-HMM test | MAE: 1.93 -> 1.93 | No clear effect |
| use v-prediction instead of x-prediction (1 step) (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 2.45 / 117.64 -> 2.12 / 133.59 | Improved |
| increase sampling from 1 to 10 steps (ablation) | Any2Any v-pred · GDP-HMM validation | MAE / Score: 2.12 / 133.59 -> 1.91 / 138.17 | Improved |
| use v-prediction instead of eps-prediction (10 steps) (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 1.93 / 137.82 -> 1.91 / 138.17 | Improved |
| fine-tune full VAE instead of frozen VAE (ablation) | Any2Any DiT · GDP-HMM validation | MAE / Score: 1.91 / 138.17 -> 2.54 / 121.76 | Hurt |
Closest papers by meaning
- Knowledge Transfer Scaling Laws for 3D Medical Imaging (NeurIPS 2026 · similarity 0.87)
- DiffusionBlend: Learning 3D Image Prior through Position-aware Diffusion Score Blending for 3D Computed Tomography Reconstruction (NeurIPS 2024 · similarity 0.85)
- How Well Do Supervised 3D Models Transfer to Medical Imaging Tasks? (ICLR 2024 · similarity 0.84)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.84)
- Addressing Asynchronicity in Clinical Multimodal Fusion via Individualized Chest X-ray Generation (NeurIPS 2024 · similarity 0.84)
- Time-to-Event Pretraining for 3D Medical Imaging (ICLR 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.
