Research map / 3D vision / 3D generation
InfBaGel: Human-Object-Scene Interaction Generation with Dynamic Perception and Iterative Refinement
Yude Zou, Junji Gong, Xing Gao, Zixuan Li, Tianxing Chen, Guanjie Zheng
This paper tried a common change: More diffusion sampling steps
increase sampling steps from 1 to 16 · HOSI benchmark · penetration vs success rate: P_mean 6.15 -> 3.39; success 87.85 -> 83.16 Mixed
What 457 accepted papers report for this change
- 277 improved
- 77 no clear effect
- 26 mixed
- 77 hurt
- Optimality of FSQ Tokens for Continuous Diffusion for Categorical Data with Application to Text-to-Speech ICML 2026 · increase denoising steps from 5 to 25 for FSQ-original · SEED-TTS test-en · WER: 2.39% -> 2.00%Improved
- DiscoForcing: A Unified Framework for Real-Time Audio-Driven Character Control with Diffusion Forcing ICML 2026 · increase denoising steps from 10 to 100 · AIST++ · latency (ms/frame): 26.26 -> 261.91Hurt
- Variable-Length Tokenization via Learnable Global Merging for Diffusion Transformers ICML 2026 · more diffusion sampling steps (25 to 100) for our method · ImageNet 256x256, 32 tokens · gFID: 2.89 -> 2.54Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 dynamic perception encoding (ablation) | InfBaGel · HOSI benchmark | success rate (%): 71.22 -> 86.35 | Improved |
| use 3 dynamic voxels instead of 1 (ablation) | InfBaGel · HOSI benchmark | success rate (%): 62.26 -> 83.16 (16 steps) | Improved |
| add bump-aware guidance on top of contact guidance (ablation) | InfBaGel · HOSI benchmark | human-scene penetration P_mean: 6.30 -> 3.39 | Improved |
| add bump-aware guidance on top of contact guidance (ablation) | InfBaGel · HOSI benchmark | human goal distance T_h (cm): 2.91 -> 4.75 | Hurt |
| add contact guidance (ablation) | InfBaGel · HOSI benchmark | contact percentage C%: 68.09 -> 79.34 | Improved |
| distill diffusion model into consistency model (ablation) | InfBaGel · HOSI benchmark | FPS: 3.38 -> 28.75 | Improved |
| distill diffusion model into consistency model (ablation) | InfBaGel · HOSI benchmark | human-scene penetration P_mean: 7.20 -> 6.38 | Improved |
| increase sampling steps from 1 to 16 (ablation) | InfBaGel (3 voxels) · HOSI benchmark | penetration vs success rate: P_mean 6.15 -> 3.39; success 87.85 -> 83.16 | Mixed |
| add LINGO HSI data to training (hybrid data) (ablation) | InfBaGel · HOSI benchmark | object-scene penetration P_mean: 16.62 -> 12.45 | Improved |
| add LINGO HSI data to training (hybrid data) (ablation) | InfBaGel · HOSI benchmark | success rate (%): 83.16 -> 81.45 | Hurt |
| raise LINGO share to 1:1 ratio (ablation) | InfBaGel · HOSI benchmark | success rate (%): 83.16 -> 69.72 | Hurt |
Closest papers by meaning
- Semantic-Aware Human Object Interaction Image Generation (ICML 2024 · similarity 0.90)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.90)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.89)
- HOI-Dyn: Learning Interaction Dynamics for Human-Object Motion Diffusion (NeurIPS 2025 · similarity 0.89)
- InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · similarity 0.88)
- HOIAnimator: Generating Text-prompt Human-object Animations using Novel Perceptive Diffusion Models (CVPR 2024 · 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.
