Research map / Reinforcement learning / Policy gradient and actor-critic
Distributional Reinforcement Learning for Efficient Exploration
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 |
|---|---|---|---|
| Replace QR-DQN-1 exploration with DLTV's decaying left-truncated variance bonus | QR-DQN-1 · Atari 49 games | average gain in cumulative rewards: 483 % | Helped |
| Use a decaying schedule rather than a constant schedule (ablation) | DLTV · Venture | cumulative rewards | Helped |
| Suppress intrinsic uncertainty with a decaying exploration-bonus schedule | QR-DQN variance-bonus exploration · 10-arm bandit counterexample | performance | Helped |
| Use left-truncated variance rather than full variance for the exploration bonus | QR-based bandit exploration · Asymmetric-reward bandit | performance | Helped |
| Use left-truncated variance rather than full variance for the exploration bonus | QR-based bandit exploration · Symmetric-reward bandit | performance | Helped |
| Select actions using q0.1 rather than the mean (ablation) | DLTV · CARLA | average distance between opposite-lane infractions: 1.35 -> 4.55 | Helped |
| Select actions using q0.1 rather than the mean (ablation) | DLTV · CARLA | average distance between car collisions: 1.53 -> 0.70 | Hurt |
| Select actions using q0.1 rather than the mean (ablation) | DLTV · CARLA | average distance between pedestrian collisions: 16.41 -> 52.33 | Helped |
| Select actions using q0.1 rather than the mean (ablation) | DLTV · CARLA | average impact of car collisions: 1078.76 -> 497.22 | Helped |
| Select actions using q0.1 rather than the mean (ablation) | DLTV · CARLA | average impact of pedestrian collisions: 40.70 -> 40.79 | Hurt |
| Select actions using q0.1 rather than the mean (ablation) | DLTV · CARLA | distance, km: 98.66 -> 104.69 | Helped |
Closest papers by meaning
- Distributional Reinforcement Learning With Quantile Regression (AAAI 2018 · similarity 0.92)
- Fully Parameterized Quantile Function for Distributional Reinforcement Learning (NeurIPS 2019 · similarity 0.91)
- Random Latent Exploration for Deep Reinforcement Learning (ICML 2024 · similarity 0.89)
- Intrinsic Benefits of Categorical Distributional Loss: Uncertainty-aware Regularized Exploration in Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Exploration and Anti-Exploration with Distributional Random Network Distillation (ICML 2024 · similarity 0.87)
- Quantile Geometry Regularization for Distributional Reinforcement Learning (NeurIPS 2026 · 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.