Research map / Reinforcement learning / Visual and representation RL
Optimistic Task Inference for Behavior Foundation Models
Thomas Rupf, Marco Bagatella, Marin Vlastelica, Andreas Krause
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 |
|---|---|---|---|
| update the task embedding every step instead of once per episode (ablation) | OpTI-BFM · DMC (Walker, Cheetah, Quadruped) | relative performance: episode-level updates lead to slower improvement than changing every step | Improved |
| use the optimistic criterion instead of Thompson sampling (ablation) | OpTI-BFM · DMC Cheetah | episodic return: significant gap between OpTI-BFM and its TS variant in Cheetah | Improved |
| use linear-structure-aware OpTI-BFM instead of black-box policy search LoLA | online task inference · DMC | episodic return: LoLA makes slower progress | Improved |
| collect data actively (OpTI-BFM, RND) instead of passive Random rollouts | task inference data source · DMC | relative performance of inferred policy | Improved |
| collect task-aware data (OpTI-BFM) instead of task-agnostic RND exploration | task inference data source · DMC | relative performance of inferred policy | Improved |
| reduce decay rate rho below 1.0 for old observations (ablation) | OpTI-BFM · DMC Walker speedup/slowdown (non-stationary) | velocity tracking: rho=1.0 struggles to adapt; reduced rho tracks target more accurately | Improved |
| make decay rate rho too small (ablation) | OpTI-BFM · DMC Walker speedup/slowdown (non-stationary) | velocity tracking | Hurt |
| warm-start with n i.i.d. labeled states (ablation) | OpTI-BFM · DMC Cheetah | return: initial performance increases quickly as n grows | Improved |
Closest papers by meaning
- Behavioral Foundation Models for Quality Diversity (NeurIPS 2026 · similarity 0.88)
- Behavior-agnostic Task Inference for Robust Offline In-context Reinforcement Learning (ICML 2025 · similarity 0.87)
- Zeroth-Order Policy Gradient for Reinforcement Learning from Human Feedback without Reward Inference (ICLR 2025 · similarity 0.87)
- Discriminative Policy Optimization for Token-Level Reward Models (ICML 2025 · similarity 0.87)
- Reward Model Learning vs. Direct Policy Optimization: A Comparative Analysis of Learning from Human Preferences (ICML 2024 · similarity 0.86)
- A Theoretical Framework for Partially-Observed Reward States in RLHF (ICLR 2025 · 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.
