Research map / Multi-agent systems and games / Human-AI and goal-conditioned
CURIOUS: Intrinsically Motivated Modular Multi-Goal 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 |
|---|---|---|---|
| replace flat HER goal space with modular goal representation (M-UVFA) (ablation) | multi-goal DDPG agent · Modular Goal Fetch Arm | success rate on achievable goals | Helped |
| single monolithic M-UVFA policy instead of one expert policy per module (MG-ME) (ablation) | M-UVFA · Modular Goal Fetch Arm | episodes to learn achievable goals: ≈450 · 103 -> ≈250 · 103 | Helped |
| active learning-progress module selection instead of random module selection (ablation) | M-UVFA · sensory failure recovery | episodes to recover 95% of performance: 78 -> 43 | Helped |
| active LP module selection vs random, recovery after sensor perturbation (ablation) | CURIOUS · sensory failure recovery | recovery time reduction: 45% faster (p < 10−4) | Helped |
| increase number of distracting modules from 0 to 4 to 7 (ablation) | CURIOUS vs M-UVFA · distracting modules | success rate on achievable goals | Helped |
| more distracting modules (ablation) | M-UVFA with random module choice · distracting modules | learning speed | Hurt |
| LP-driven attention detects competence drop on forgotten module M3 | CURIOUS single run · Modular Goal Fetch Arm, 4 modules | module selection probability | Helped |
Closest papers by meaning
- Automatic Curriculum Learning through Value Disagreement (NeurIPS 2020 · similarity 0.90)
- Automated curriculum generation through setter-solver interactions (ICLR 2020 · similarity 0.88)
- Maximum Entropy Gain Exploration for Long Horizon Multi-goal Reinforcement Learning (ICML 2020 · similarity 0.88)
- Wonder Wins Ways: Curiosity-Driven Exploration through Multi-Agent Contextual Calibration (NeurIPS 2025 · similarity 0.86)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals (ICLR 2025 · similarity 0.86)
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.