Research map / Reinforcement learning / Visual and representation RL
Compositional Reinforcement Learning from Logical Specifications
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| decompose the specification into an abstract graph and interleave Dijkstra-style planning with per-edge RL (DiRL) versus monolithic reward-based baselines | DiRL · 16-Rooms environment, specifications phi_3 to phi_5 | probability of satisfying the specification | Helped |
| DiRL versus Spectrl, Qrm, Hrm, Tltl and R-avi baselines | DiRL · Fetch PickAndPlace, PickAndPlaceStatic, PickAndPlaceChoice | probability of success | Helped |
| increase specification size (abstract graph size) | DiRL · 16-Rooms environment | samples to reach success probability z | No clear effect |
| train each edge policy at most once lazily instead of retraining edge policies repeatedly with different initial distributions (R-avi) | DiRL · 16-Rooms environment | probability of success | Helped |
| block many doors in 16-Rooms | DiRL · 16-Rooms with blocked doors | probability of success | No clear effect |
Closest papers by meaning
- Run-Time Task Composition with Safety Semantics (ICML 2024 · similarity 0.87)
- Skill Machines: Temporal Logic Skill Composition in Reinforcement Learning (ICLR 2024 · similarity 0.87)
- Reconciling Spatial and Temporal Abstractions for Goal Representation (ICLR 2024 · similarity 0.87)
- Learning Planning Abstractions from Language (ICLR 2024 · similarity 0.87)
- Compositional Automata Embeddings for Goal-Conditioned Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Compositional Reasoning in Language Models under Reinforcement Learning Post-Training (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.