Research map / Reinforcement learning / Visual and representation RL
Regularized Latent Dynamics Prediction is a Strong Baseline For Behavioral Foundation Models
Pranaya Jajoo, Harshit Sikchi, Siddhant Agarwal, Amy Zhang, Scott Niekum, Martha White
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 orthogonality regularization (lambda>0) to latent dynamics prediction (ablation) | RLDP · zero-shot RL (DMC) | average return | Improved |
| add spherical normalization on predicted latent next state g (ablation) | RLDP · Quadruped (DMC) | average return: 514.02 -> 612.77 | Improved |
| add spherical normalization on g (ablation) | RLDP · Pointmass (DMC) | average return: 581.63 -> 671.11 | Improved |
| add spherical normalization on g (ablation) | RLDP · Quadruped Walk (DMC) | return: 465.40 vs 519.39 without SN | Hurt |
| learned representations (FB, PSM, RLDP) instead of random features (ablation) | zero-shot BFM · ExoRL DMC | return: e.g. Walker Stand 392.40 -> 890.40 | Improved |
| RLDP vs FB and PSM in offline DMC zero-shot | RLDP · ExoRL DMC (Walker, Cheetah, Quadruped, Pointmass) | return: Walker Flip 977.08 (FB) vs 492.94; Quadruped Jump 549.57 (PSM) vs 733.32 | Mixed |
| RLDP representations in low-coverage D4RL | RLDP · D4RL (6 tasks) | normalized return: outperforms all baselines in 5 of 6 tasks | Improved |
| explicit Bellman backups in representation learning (ablation) | RLDP · low-coverage D4RL | policy performance | Hurt |
| RLDP on SMPL Humanoid (45 tasks) | RLDP · SMPL Humanoid | normalized return | No clear effect |
Closest papers by meaning
- Foundation Policies with Hilbert Representations (ICML 2024 · similarity 0.88)
- Towards Robust Zero-Shot Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Towards Principled Representation Learning from Videos for Reinforcement Learning (ICLR 2024 · similarity 0.87)
- Distributional Successor Features Enable Zero-Shot Policy Optimization (NeurIPS 2024 · similarity 0.86)
- Behavioral Foundation Models for Quality Diversity (NeurIPS 2026 · similarity 0.86)
- Bridging State and History Representations: Understanding Self-Predictive RL (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.
