Research map / Multi-agent systems and games / Human-AI and goal-conditioned
An Asymmetric Modeling for Action Assessment
Jibin Gao, Wei-Shi Zheng, Jia-Hui Pan, Chengying Gao, Yaowei Wang, Wei Zeng, Jianhuang Lai
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 the asymmetric interaction module (AIM) to the I3D baseline (ablation) | Ours · JIGSAWS (4-fold CV) | Spearman correlation (%): 8 -> 71 | Improved |
| add attention fusion instead of fusion in each half (ablation) | Ours · JIGSAWS (4-fold CV) | Spearman correlation (%): 67 -> 71 | Improved |
| add the transformation module (ablation) | Ours · JIGSAWS (4-fold CV) | Spearman correlation (%): 68 -> 71 | Improved |
| add the difference module (ablation) | Ours · JIGSAWS (4-fold CV) | Spearman correlation (%): 68 -> 71 | Improved |
| assign primary and secondary roles correctly rather than exchanging them (ablation) | Ours · JIGSAWS (4-fold CV) | Spearman correlation (%): 67 -> 71 | Improved |
| exchange primary and secondary on weak-asymmetry actions (ablation) | Ours · TASD-2 SyncDiving-3m | Spearman correlation (%): 91.50 -> 91.75 | No clear effect |
| exchange primary and secondary on weak-asymmetry actions (ablation) | Ours · TASD-2 SyncDiving-10m | Spearman correlation (%): 85.13 -> 85.10 | No clear effect |
| use multi-task training instead of single-task | Ours · TASD-2 Average | Spearman correlation (%): 87 -> 89 | Improved |
| add AIM on top of I3D+Kinetics features | Baseline+Kinetic · JIGSAWS (4-fold CV) | Spearman correlation (%): 46 -> 71 | Improved |
| add pose features to the baseline | Baseline · TASD-2 Average | Spearman correlation (%): 82 -> 84 | Improved |
| add AIM to the baseline on TASD-2 (single-task) | Baseline · TASD-2 Average | Spearman correlation (%): 82 -> 87 | Improved |
| apply method to single-person actions with camera motion as secondary agent | Ours vs JR-GCN · AQA-7 Average | Spearman correlation (%): 78.49 -> 77.89 | No clear effect |
Closest papers by meaning
- Adaptive Action Assessment (TPAMI 2021 · similarity 0.86)
- Role-Aware Interaction Generation from Textual Description (ICCV 2023 · similarity 0.86)
- InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · similarity 0.82)
- ReGenNet: Towards Human Action-Reaction Synthesis (CVPR 2024 · similarity 0.82)
- AgentStudio: A Toolkit for Building General Virtual Agents (ICLR 2025 · similarity 0.82)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.82)
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.
