Research map / Reinforcement learning / Visual and representation RL
PvP: Data-Efficient Humanoid Robot Learning with Proprioceptive-Privileged Contrastive Representations
Mingqi Yuan, Tao Yu, Haolin Song, Bo Li, Xin Jin, Hua Chen, Wenjun Zeng
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 PvP proprioceptive-privileged contrastive loss to PPO policy encoder | PPO · LimX-Oli-31dof-Velocity | normalized score (learning speed) | Improved |
| add PvP loss to PPO policy encoder | PPO · LimX-Oli-31dof-Mimic | normalized score | Improved |
| add VAE reconstruction loss to PPO | PPO · LimX-Oli-31dof-Mimic | normalized score | Hurt |
| add PvP to PPO | PPO · LimX-Oli-31dof-Velocity | action smoothness penalty | Improved |
| add PvP to PPO | PPO · LimX-Oli-31dof-Mimic | tracking (waist pitch, feet distance, joint position) | Improved |
| update SRL loss every 50 steps instead of every step or every 100 (ablation) | SRL methods (VAE, SPR, SimSiam, PvP) · LimX-Oli-31dof-Mimic | normalized score | Improved |
| vary SRL update interval (1, 50, 100) (ablation) | SRL methods · LimX-Oli-31dof-Velocity | normalized score | No clear effect |
| use more SRL training data (10% -> 50% -> 100% of rollouts) (ablation) | SimSiam and PvP · LimX-Oli-31dof-Mimic | normalized score | Improved |
| vary SRL training data proportion (ablation) | SRL methods · LimX-Oli-31dof-Velocity | normalized score | No clear effect |
| apply SRL loss to value encoder instead of policy encoder (ablation) | SimSiam and VAE with PPO · LimX-Oli-31dof-Velocity and Mimic | convergence speed / stability | Hurt |
Closest papers by meaning
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.89)
- Hierarchical World Models as Visual Whole-Body Humanoid Controllers (ICLR 2025 · similarity 0.87)
- Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · similarity 0.87)
- Zero-Shot Whole-Body Humanoid Control via Behavioral Foundation Models (ICLR 2025 · similarity 0.86)
- HumanoidArena: Benchmarking Egocentric Hierarchical Whole-body Learning (NeurIPS 2026 · similarity 0.86)
- SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework (NeurIPS 2026 · 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.
