Research map / Reinforcement learning / Visual and representation RL
Unsupervised Control Through Non-Parametric Discriminative Rewards
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 |
|---|---|---|---|
| Use hindsight experience replay only, without a learned goal achievement reward (ablation) | DISCERN · Seaquest | goals achieved: 12% | Hurt |
| Use hindsight experience replay only, without a learned goal achievement reward (ablation) | DISCERN · Montezuma's Revenge | goals achieved: 11.4% | Hurt |
| Remove hindsight experience replay (ablation) | DISCERN · Seaquest | goals achieved: 11.4% | Hurt |
| Remove hindsight experience replay (ablation) | DISCERN · Montezuma's Revenge | goals achieved: 8% | Hurt |
| Use a learned discriminative reward rather than pixel distance, conditioned autoencoder, or conditioned WGAN reward | DISCERN · Seaquest and Montezuma's Revenge | goal achievement | Helped |
| Use a learned discriminative reward rather than baseline reward methods | DISCERN · DeepMind Control Suite | goal achievement | Mixed |
| Use a learned discriminative reward rather than baseline reward methods | DISCERN · DeepMind Control Suite: cartpole and finger | controllable state dimensions matched | Helped |
| Use a learned discriminative reward rather than a conditioned autoencoder | DISCERN · DeepMind Control Suite: manipulator | controllable state dimensions matched | Helped |
Closest papers by meaning
- Learn Goal-Conditioned Policy with Intrinsic Motivation for Deep Reinforcement Learning (AAAI 2022 · similarity 0.90)
- Visual Reinforcement Learning with Imagined Goals (NeurIPS 2018 · similarity 0.90)
- Reinforcement Learning with Unsupervised Auxiliary Tasks (ICLR 2017 · similarity 0.90)
- Towards Principled Unsupervised Multi-Agent Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Foundation Policies with Hilbert Representations (ICML 2024 · similarity 0.87)
- Unifying Goal-Conditioned RL and Unsupervised Skill Learning via Control-Maximization (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.