Research map / Reinforcement learning / Visual and representation RL
Reinforcement Learning with Unsupervised Auxiliary Tasks
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 |
|---|---|---|---|
| Add pixel control to A3C (ablation) | A3C · Labyrinth | performance | Helped |
| Add reward prediction to A3C (ablation) | A3C · Labyrinth | performance | Helped |
| Add value function replay to A3C (ablation) | A3C · Labyrinth | performance | Helped |
| Combine pixel control, reward prediction, and value function replay | A3C · Labyrinth (13 levels) | mean human-normalised score: 54% -> 87% | Helped |
| Combine pixel control, reward prediction, and value function replay | A3C · Labyrinth (13 levels) | mean speedup in steps to reach A3C best performance: 10× | Helped |
| Add auxiliary tasks to A3C | A3C · Atari (57 games) | mean human-normalised performance: 880% | Helped |
| Add pixel control to A3C | A3C · Labyrinth navigation | final performance | Helped |
| Predict immediate input changes instead of reconstructing the input | A3C self-supervised learning · Labyrinth navigation | performance | Helped |
| Add input reconstruction loss | A3C · Labyrinth navigation | final score | Hurt |
| Add input reconstruction loss | A3C · Labyrinth navigation | initial learning speed | Helped |
| Add feature control | A3C · Labyrinth | performance | Helped |
| Remove pixel control from UNREAL (ablation) | UNREAL · Atari (57 games) | human-normalised performance | Hurt |
Closest papers by meaning
- Unsupervised Control Through Non-Parametric Discriminative Rewards (ICLR 2019 · similarity 0.90)
- Exploratory Diffusion Model for Unsupervised Reinforcement Learning (ICLR 2026 · similarity 0.89)
- Behavior From the Void: Unsupervised Active Pre-Training (NeurIPS 2021 · similarity 0.88)
- Never Give Up: Learning Directed Exploration Strategies (ICLR 2020 · similarity 0.88)
- Unsupervised Domain Adaptation with Dynamics-Aware Rewards in Reinforcement Learning (NeurIPS 2021 · similarity 0.88)
- To the Max: Reinventing Reward in Reinforcement Learning (ICML 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.