Research map / Video / Video generation
Towards Highly-Constrained Human Motion Generation with Retrieval-Guided Diffusion Noise Optimization
Hanchao Liu, Fang-Lue Zhang, Shining Zhang, Tai-Jiang Mu, Shi-Min Hu
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 retrieval-guided noise initialization to DNO (narrow gap) | ProgMoGen+DNO · Task-1 very narrow gap | Constraint error: 0.0162 -> 0.0050 | Improved |
| add retrieval-guided noise initialization to DNO (low barrier) | ProgMoGen+DNO · Task-2 very low barrier | Constraint error: 0.000115 -> 0.000049 | Improved |
| add retrieval-guided noise initialization to DNO (walking steps) | ProgMoGen+DNO · Task-3 walking steps with raise hand | Success rate: 0.469 -> 0.594 | Improved |
| add retrieval guidance to DNO on foot skate (narrow gap) | ProgMoGen+DNO · Task-1 very narrow gap | Foot skate: 0.029 -> 0.040 | Hurt |
| add retrieval guidance to DNO to reduce jitter (low barrier) | ProgMoGen+DNO · Task-2 very low barrier | Max joint acceleration: 0.261 -> 0.194 | Improved |
| add 5-run initial noise search to the retrieval method | Ours · Task-1 very narrow gap | Constraint error: 0.0050 -> 0.0000 | Improved |
| add retrieved noise z2 to random-only noise z1 (ablation) | Ours · Task-2 very low barrier ablation | Constraint error: 0.000115 -> 0.000049 | Improved |
| add random noise z1 to retrieval-only noise z2 (ablation) | Ours · Task-2 very low barrier ablation | Local foot skate: 0.180 -> 0.134 | Improved |
| add relational task parsing C_R (ablation) | Ours · Task-2 very low barrier ablation | Constraint error: 0.000132 -> 0.000049 | Improved |
| add optimized mask M instead of fixed 0.5 weighting (ablation) | Ours · Task-2 very low barrier ablation | Local foot skate: 0.306 -> 0.134 | Improved |
| add motion-quality reward function (ablation) | Ours · Task-2 very low barrier ablation | Max joint acceleration: 0.202 -> 0.194 | Improved |
| split constraints into C1 and C2 instead of C1=C2=C (ablation) | Ours · Task-2 very low barrier ablation | Constraint error: 0.000052 -> 0.000049 | Mixed |
Closest papers by meaning
- Optimizing Diffusion Noise Can Serve As Universal Motion Priors (CVPR 2024 · similarity 0.90)
- Rethinking Diffusion for Text-Driven Human Motion Generation: Redundant Representations, Evaluation, and Masked Autoregression (CVPR 2025 · similarity 0.88)
- CoDA: Coordinated Diffusion Noise Optimization for Whole-Body Manipulation of Articulated Objects (NeurIPS 2025 · similarity 0.88)
- Seamless Human Motion Composition with Blended Positional Encodings (CVPR 2024 · similarity 0.88)
- Human Motion Diffusion as a Generative Prior (ICLR 2024 · similarity 0.87)
- Unified Noise Steering for Efficient Human-Guided VLA Adaptation (NeurIPS 2026 · 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.
