Research map / Reinforcement learning / Visual and representation RL
What Reward Structure Enables Efficient Sparse-Reward RL? A Proof-of-Concept with Policy-Aware Matrix Completion
Ibne Farabi Shihab, Sanjeda Akter, Anuj Sharma
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 inverse propensity weighting (ablation) | PAMC · Atari (26 games) | human-normalized score: 1.15 -> 1.42 | Improved |
| add PAMC reward completion to DrQ-v2 | DrQ-v2 · Montezuma's Revenge (10M steps) | return: 200 ± 50 -> 4100 ± 250 | Improved |
| add PAMC to DrQ-v2 | DrQ-v2 · Atari (26 games) | human-normalized score: 1.25 ± 0.05 -> 1.51 ± 0.05 | Improved |
| add PAMC to MT-SAC | MT-SAC · MetaWorld-50 | success rate: 0.65 ± 0.03 -> 0.78 ± 0.02 | Improved |
| add PAMC to DreamerV3 | DreamerV3 · DM Control | return: 820 ± 16 -> 921 ± 17 | Improved |
| apply PAMC on dense-reward Walker-Walk | PAMC · Walker-Walk | return: 920 ± 30 -> 950 ± 25 | No clear effect |
| apply PAMC on high-rank hashed reward controls (ablation) | PAMC · high-rank hash-reward | return | No clear effect |
| reduce rank by coarse discretization (ablation) | PAMC · reward matrix | return | No clear effect |
| apply PAMC to D4RL MuJoCo with CQL | CQL · D4RL MuJoCo | return: +15% | Improved |
| corrupt propensities with sigma 0.3 noise (ablation) | PAMC · Atari HNS | HNS: -8% | 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.
- Agent57 — Agent57: Outperforming the Atari Human Benchmark (ICML 2020) · better on 2 of 2 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
- Addressing Sparse-Rewards in RL with Scalable Hierarchical Novel Eigen Options (NeurIPS 2026 · similarity 0.91)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.90)
- Learning Process Rewards via Visitation Matching for Efficient RL (NeurIPS 2026 · similarity 0.88)
- Action-Dependent Optimality-Preserving Reward Shaping (ICML 2025 · similarity 0.88)
- A Geometric Perspective on Reward Function Updates in Inverse Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Catching Two Birds with One Stone: Reward Shaping with Dual Random Networks for Balancing Exploration and Exploitation (ICML 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.
