Research map / Reinforcement learning / Visual and representation RL
Memory Based Trajectory-conditioned Policies for Learning from Sparse Rewards
Yijie Guo, Jongwook Choi, Marcin Moczulski, Shengyu Feng, Samy Bengio, Mohammad Norouzi, Honglak Lee
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 diverse trajectory-conditioned imitation instead of count-based exploration alone | PPO+EXP · Montezuma's Revenge | score: 12,338 -> 22,616 | Improved |
| Use diverse trajectory-conditioned imitation instead of count-based exploration alone | PPO+EXP · Pitfall | score: 0 -> 12,446 | Improved |
| Use diverse trajectory-conditioned imitation instead of count-based exploration alone | PPO+EXP · Venture | score: 1,817 -> 2,011 | Improved |
| Use diverse trajectory-conditioned imitation instead of high-reward self-imitation | PPO+SIL · Apple-Gold | average episode reward: 2 -> 8.5 | Improved |
| Use diverse trajectory-conditioned imitation instead of count-based exploration alone | PPO+EXP · Apple-Gold | average episode reward: 4 -> 8.5 | Improved |
| Add a count-based exploration bonus to DTSIL (ablation) | DTSIL · Atari hard-exploration games | exploration performance | Improved |
| Replace flexible trajectory-conditioned actions with repeated stored actions (ablation) | DTRA · Atari stochastic environments | performance | Hurt |
| Condition the policy on observation sequences rather than learned state-embedding sequences (ablation) | DTSIL · Montezuma's Revenge with learned state representations | score: >20,000 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Agent57 — Agent57: Outperforming the Atari Human Benchmark (ICML 2020) · better on 0 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
- Search on the Replay Buffer: Bridging Planning and Reinforcement Learning (NeurIPS 2019 · similarity 0.89)
- Reparameterized Policy Learning for Multimodal Trajectory Optimization (ICML 2023 · similarity 0.88)
- Action-Dependent Optimality-Preserving Reward Shaping (ICML 2025 · similarity 0.88)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Learning Process Rewards via Visitation Matching for Efficient RL (NeurIPS 2026 · similarity 0.87)
- DreamSmooth: Improving Model-based Reinforcement Learning via Reward Smoothing (ICLR 2024 · 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.
