Research map / Reinforcement learning / Visual and representation RL
Compositional Transduction with Latent Analogies for Offline Goal-Conditioned Reinforcement Learning
Junseok Kim, Dohyeong Kim, Mineui Hong, Songhwai Oh
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 bilinear transduction and dual analogy value/policies (CTA) on top of HIQL∨+α∨ (ablation) | CTA · OGBench manipulation average | success rate (%): 39.6 -> 56.3 | Improved |
| replace dual goal representation φ(g) with dual analogy α∨(s,g) in HIQL∨, without bilinear transduction (ablation) | HIQL∨ · OGBench manipulation average | success rate (%): 39.3 -> 39.6 | No clear effect |
| CTA versus dual-goal-representation baselines | CTA · OGBench manipulation average | success rate (%): 39.3 -> 56.3 (HIQL∨) | Improved |
| CTA versus HIQL∨+α∨ on puzzle-4x4 (ablation) | CTA · OGBench puzzle-4x4-play | success rate (%): 50±5 -> 84±3 | Improved |
| CTA versus GCIVL∨ on cube-double | CTA · OGBench cube-double-play | success rate (%): 60±4 -> 50±5 | Hurt |
| CTA versus GCIQL on puzzle-3x3 | CTA · OGBench puzzle-3x3-play | success rate (%): 95±1 vs 94±11 | No clear effect |
| CTA versus best baselines on puzzle-4x5 and 4x6 | CTA · OGBench puzzle-4x5/4x6-play | success rate (%): 14±1 -> 17±1 (4x5); 12±1 -> 12±2 (4x6) | No clear effect |
| hold out OOC context-task pairs; CTA vs HIQL∨+α∨ on scene (ablation) | CTA · scene-play direct OOC case study | direct success rate (%): 48±14 -> 73±9 | Improved |
| hold out OOC context-task pairs; CTA vs HIQL∨+α∨ on puzzle-4x4 (ablation) | CTA · puzzle-4x4-play direct OOC case study | direct success rate (%): 66±11 -> 80±8 | Improved |
| use dual analogy only without bilinear transduction under OOC holdout (HIQL∨ vs HIQL∨+α∨) (ablation) | HIQL∨ · scene-play direct OOC case study | direct success rate (%): 45±11 -> 48±14 | No clear effect |
| use dual analogy in HIQL∨ on puzzle-4x4 OOC holdout (ablation) | HIQL∨ · puzzle-4x4-play direct OOC case study | direct success rate (%): 35±17 -> 66±11 | Improved |
| CTA has about 20% fewer parameters than HIQL∨+α∨ yet performs better | CTA · OGBench manipulation | success rate (%): about 20% fewer parameters | Improved |
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 7 of 8 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
- Modeling Unseen Environments with Language-guided Composable Causal Components in Reinforcement Learning (ICLR 2025 · similarity 0.89)
- Compositional Reasoning in Language Models under Reinforcement Learning Post-Training (NeurIPS 2026 · similarity 0.88)
- Pessimistic Latent Task-aware Optimization for Robust Offline Meta-Reinforcement Learning (NeurIPS 2026 · similarity 0.87)
- Horizon Generalization in Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Temporal Representation Alignment: Successor Features Enable Emergent Compositionality in Robot Instruction Following (NeurIPS 2025 · similarity 0.87)
- A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals (ICLR 2025 · similarity 0.87)
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.
