Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Curiosity-Driven Exploration via Latent Bayesian Surprise
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 latent-space Bayesian surprise rather than parameter-space Bayesian surprise | LBS exploration · Mountain Car | state-space coverage: around 90% for LBS, VIME, ICM and Disagreement | No clear effect |
| Use latent-space Bayesian surprise rather than parameter-space Bayesian surprise | LBS exploration · Half-Cheetah | state-space coverage: around 73% vs $\sim$49% | Helped |
| Use latent-space Bayesian surprise rather than prediction error to generate curiosity rewards | LBS · Stochastic Mountain Car | state-space coverage | Helped |
| Use latent Bayesian surprise to score transitions | curiosity-driven exploration · Stochastic MNIST | intrinsic motivation ratio at convergence | Helped |
| Use latent Bayesian surprise instead of pixel-space surprisal | arcade-game exploration · Arcade games | average game score: best average final score in 5 out of 9 games | Helped |
| Reconstruct features instead of pixels (ablation) | LBS · Arcade games | game score | Helped |
| Remove episodic-count modulation (ablation) | NGU and RIDE · Continuous control and stochastic benchmarks | exploration performance | Hurt |
| Use pixel reconstruction instead of feature reconstruction (ablation) | LBS · Arcade games | game score | Hurt |
| Add episodic-count modulation (ablation) | NGU and RIDE · Stochastic benchmarks | exploration performance | Helped |
Closest papers by meaning
- Curiosity in Hindsight: Intrinsic Exploration in Stochastic Environments (ICML 2023 · similarity 0.91)
- Curiosity-driven Exploration by Self-supervised Prediction (ICML 2017 · similarity 0.89)
- Wonder Wins Ways: Curiosity-Driven Exploration through Multi-Agent Contextual Calibration (NeurIPS 2025 · similarity 0.89)
- Bayesian Exploration Networks (ICML 2024 · similarity 0.87)
- Beyond Optimism: Exploration With Partially Observable Rewards (NeurIPS 2024 · similarity 0.85)
- Reinforcement Learning Under Latent Dynamics: Toward Statistical and Algorithmic Modularity (NeurIPS 2024 · similarity 0.85)
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.