Research map / Video / Video generation
PRIMAL: Physically Reactive and Interactive Motor Model for Avatar Learning
Yan Zhang, Yao Feng, Alpár Cseke, Nitin Saini, Nathan Bajandas, Nicolas Heron, Michael J. Black
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 in-context conditioning instead of AdaLN (ablation) | PRIMAL · HumanEva/SFU motion realism | ARD x1e-3: 0.147 -> 0.074 | Improved |
| predict 15 frames rather than 8 frames (ablation) | PRIMAL InContext · SFU | ASR: 0.073 -> 0.027 | Improved |
| add Gaussian noise to joint locations and velocities (ablation) | PRIMAL InContext · HumanEva/SFU | ASR/ARD: 0.027 -> 0.022 ASR (SFU); ARD 0.081 -> 0.098 | No clear effect |
| replace SiLU with ReLU (ablation) | PRIMAL InContext · SFU | ASR: 0.027 -> 0.042 | Hurt |
| increase attention latent dim from 256 to 512 (ablation) | PRIMAL InContext · SFU | ARD x1e-3: 0.081 -> 0.105 | No clear effect |
| add inertialization at test time (ablation) | PRIMAL InContext · HumanEva ASR | ASR: 0.022 -> 0.015 | Improved |
| add joint re-projection at test time (ablation) | PRIMAL InContext · HumanEva ASR | ASR: 0.022 -> 0.034 | Hurt |
| use separate-token conditioning (CAMDM style) (ablation) | PRIMAL InContext · HumanEva | ASR: 0.023 -> 0.037 | Hurt |
| use ControlNet-based adaptor instead of direct finetuning | PRIMAL · Few-shot action generation | R-precision: 0.87 -> 0.96 | Improved |
| use proposed adaptor instead of OmniControlNet for spatial target reaching | PRIMAL · Spatial target reaching | err. dist.: 0.106 -> 0.031; ASR 0.249 -> 0.338 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- CLoSD — CLoSD: Closing the Loop between Simulation and Diffusion for multi-task character control (ICLR 2025) · better on 1 of 2 shared results
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
- AvatarGO: Zero-shot 4D Human-Object Interaction Generation and Animation (ICLR 2025 · similarity 0.89)
- RAM-Avatar: Real-time Photo-Realistic Avatar from Monocular Videos with Full-body Control (CVPR 2024 · similarity 0.89)
- Ready-to-React: Online Reaction Policy for Two-Character Interaction Generation (ICLR 2025 · similarity 0.87)
- Vid2Avatar-Pro: Authentic Avatar from Videos in the Wild via Universal Prior (CVPR 2025 · similarity 0.87)
- MPMAvatar: Learning 3D Gaussian Avatars with Accurate and Robust Physics-Based Dynamics (NeurIPS 2025 · similarity 0.86)
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.86)
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.
