Research map / Video / Video segmentation and tracking
Learning Transferable Interaction Primitives from Game Videos for Humanoid Locomotion
Xiangming Zhu, Huayu Deng, Haoran Zhao, Yiwei Hao, Yunbo Wang
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 TRIP instead of PULSE | humanoid policy · Reach (1.5x terrain) | performance score: 97±3.7 -> 123±8.2 | Improved |
| use TRIP instead of PULSE | humanoid policy · Reach (unseen wave terrain) | performance score: 114±2.3 -> 183±1.8 | Improved |
| use TRIP instead of PULSE | humanoid policy · Trajectory (external disturbances) | performance score: 176±7.3 -> 203±8.3 | Improved |
| use TRIP instead of PULSE+AMP | humanoid policy · Speed (1.5x terrain) | performance score: 180±2.8 -> 216±3.4 | Improved |
| add interaction primitives (ablation) | TRIP · Trajectory | tracking error and success rate | Improved |
| add cross-domain context alignment (ablation) | TRIP · Trajectory | tracking error and success rate | Improved |
| add mutual (two-sided) regularization Eq. 11 (ablation) | TRIP · Trajectory | tracking error and success rate | Improved |
| use two-sided instead of one-sided regularization (ablation) | TRIP · Trajectory | tracking error and success rate | Improved |
| use inferred posterior context instead of random context (ablation) | predictor · next-state prediction | State MSE (x10^-3): 8.18 -> 4.48 | Improved |
| use inferred posterior context instead of random context (ablation) | predictor · action prediction | Action MSE (x10^-3): 0.869 -> 0.581 | Improved |
| increase codebook size from 64 to 256 (ablation) | TRIP · codebook ablation | reconstruction error: 0.397 -> 0.265 | Improved |
| increase codebook size from 256 to 512 (ablation) | TRIP · codebook ablation | reconstruction error and dead codes: 0.265 -> 0.255; dead codes 4 -> 52 | No clear effect |
Closest papers by meaning
- SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework (NeurIPS 2026 · similarity 0.89)
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.88)
- DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control (NeurIPS 2026 · similarity 0.88)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.87)
- Grounding Video Models to Actions through Goal Conditioned Exploration (ICLR 2025 · similarity 0.87)
- Humanoid Locomotion as Next Token Prediction (NeurIPS 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.
