Research map / Reinforcement learning / Visual and representation RL
Learn Goal-Conditioned Policy with Intrinsic Motivation for Deep Reinforcement Learning
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 GPIM instead of RIG, DISCERN, L2 Distance, or RIG+ | goal-conditioned policy learning · 2D navigation (x-y goal) | normalized distance to goals | Mixed |
| use GPIM instead of the baselines | goal-conditioned policy learning · 2D navigation (color-shape goal) and object manipulation | normalized distance to goals | Helped |
| use GPIM instead of the baselines | goal-conditioned policy learning · Atari and MuJoCo tasks | normalized distance to goals | Helped |
| use GPIM instead of RIG and DISCERN with the same exploration-generated goals | goal-conditioned policy learning · maze task | learning performance | Helped |
| use GPIM instead of DIAYN-Imitator and DIAYN-Goal | goal-conditioned policy learning · 2D navigation | goal-reaching performance | Helped |
| remove the policy-loss gradients updating the goal encoder (ablation) | GPIM · 2D navigation | learning performance | Hurt |
| remove the KL prior-matching loss (ablation) | GPIM · 2D navigation | goal disentanglement | Hurt |
| remove the reconstruction loss (ablation) | GPIM · 2D navigation | learning performance | No clear effect |
| remove the reconstruction loss (ablation) | GPIM · MuJoCo tasks | learning performance | Hurt |
| remove the policy-loss gradients updating the goal encoder (ablation) | GPIM · MuJoCo tasks | learning performance | Hurt |
| remove the KL prior-matching loss (ablation) | GPIM · MuJoCo tasks | learning performance | No clear effect |
| increase discrete skill count from 5 to 100 (ablation) | GPIM · 2D Nav. | normalized distance to goals: 0.296 (+/- 0.091) -> 0.132 (+/- 0.037) | Helped |
Closest papers by meaning
- Visual Reinforcement Learning with Imagined Goals (NeurIPS 2018 · similarity 0.90)
- Unsupervised Control Through Non-Parametric Discriminative Rewards (ICLR 2019 · similarity 0.90)
- PcLast: Discovering Plannable Continuous Latent States (ICML 2024 · similarity 0.88)
- Flattening Hierarchies with Policy Bootstrapping (NeurIPS 2025 · similarity 0.88)
- Learning Goal-Conditioned Representations for Language Reward Models (NeurIPS 2024 · similarity 0.88)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · similarity 0.88)
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.