Research map / Reinforcement learning / Visual and representation RL
Laplacian Representations for Decision-Time Planning
Dikshant Shehmar, Matthew Kyle Schlegel, Matthew E. Taylor, Marlos C. Machado
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 scaled Laplacian latent space instead of PcLast contrastive latent (no prior) | ALPS · Maze2D-PointMass Spiral | success rate (%): 35 -> 91 (1 cluster); 60 -> 94 (16 clusters) | Improved |
| use scaled Laplacian latent space instead of PcLast latent | ALPS · Maze2D-PointMass Rooms | success rate (%): 30 -> 92 (1 cluster) | Improved |
| add high-level subgoal planner (16 clusters vs 1) to PcLast (ablation) | PcLast · Maze2D-PointMass Spiral | success rate (%): 35 -> 60 | Improved |
| add high-level subgoal planner to ALPS (ablation) | ALPS† · Maze2D-PointMass Spiral | success rate (%): 91 -> 94 | No clear effect |
| add hierarchical Dijkstra subgoal planner (ablation) | ALPS · OGBench antmaze-large-explore | success rate (%): 1 -> 43 (pi_prior alone vs Dijkstra + pi_prior) | Improved |
| add behavior prior to CEM (ablation) | ALPS · OGBench antmaze-large-navigate | success rate (%): 0 -> 93 (Dijkstra+CEM vs ALPS) | Improved |
| add behavior prior to CEM (ablation) | ALPS · OGBench pointmaze-large-navigate | success rate (%): 100 -> 80 | Hurt |
| add CEM on top of behavior prior (ablation) | ALPS · OGBench antmaze-large-explore | success rate (%): 43 -> 90 | Improved |
| increase number of clusters in high-level planner (ablation) | ALPS · OGBench antmaze-large (varying Kmax/explore data) | success rate | Improved |
| use ALPS instead of model-free baselines on giant mazes | ALPS · OGBench giant mazes | success rate (%): humanoidmaze-giant-navigate: HIQL 12 -> ALPS 67 | Improved |
| use ALPS on teleport mazes | ALPS · OGBench teleport mazes | success rate (%): pointmaze-teleport-stitch: HIQL 34 vs ALPS 13 | Mixed |
| use ALPS on multi-cube manipulation | ALPS · OGBench cube-double-play | success rate (%): 68 (single) -> 2 (double) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HIQL — HIQL: Offline Goal-Conditioned RL with Latent States as Actions (NeurIPS 2023) · better on 10 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Online Laplacian-Based Representation Learning in Reinforcement Learning (ICML 2025 · similarity 0.90)
- PcLast: Discovering Plannable Continuous Latent States (ICML 2024 · similarity 0.89)
- Proper Laplacian Representation Learning (ICLR 2024 · similarity 0.88)
- Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Hitting Time Isomorphism for Multi-Stage Planning with Foundation Policies (NeurIPS 2026 · similarity 0.88)
- Reconciling Spatial and Temporal Abstractions for Goal Representation (ICLR 2024 · similarity 0.88)
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.
