Research map / Probabilistic and scientific modeling / Molecules and proteins
Deep active inference agents using Monte-Carlo methods
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 DAIMC instead of DQN or A2C | Animal-AI preferences · Animal-AI | task performance | Helped |
| Use DAIMC instead of PPO2 | Animal-AI preferences · Animal-AI | task performance | No clear effect |
| Use the full active-inference agent rather than the reward-maximizing baseline | Dynamic dSprites object sorting · Dynamic dSprites | reward collection | No clear effect |
| Use active-inference EFE formulations rather than reward maximization alone | Dynamic dSprites object sorting · Dynamic dSprites | reward reconstruction loss | Helped |
| Add top-down attention through the state precision factor (ablation) | DAIMC latent state representation · Dynamic dSprites | total correlation | Helped |
| Remove parameter-belief uncertainty from EFE, using latent-state beliefs only (ablation) | DAIMC · Dynamic dSprites object sorting | reward collection | No clear effect |
| Down-weight the reward signal in the EFE functional (ablation) | DAIMC · Dynamic dSprites object sorting | reward collection | No clear effect |
| Add KL regularization toward a standard normal latent prior (ablation) | DAIMC encoder and observation networks · Training | speed of convergence | Helped |
| Add KL regularization toward a standard normal latent prior (ablation) | DAIMC · Training behavior | agent behavior | No clear effect |
| Increase the decision threshold (ablation) | DAIMC MCTS planner · Dynamic dSprites | task performance: T_dec ≈ 0.8 | Mixed |
| Use shorter decision thresholds during visual input removal (ablation) | DAIMC · Animal-AI lights-off | performance over time without visual input | Helped |
Closest papers by meaning
- ExACT: Teaching AI Agents to Explore with Reflective-MCTS and Exploratory Learning (ICLR 2025 · similarity 0.86)
- Evolving Minds: Logic-Informed Inference from Temporal Action Patterns (ICML 2025 · similarity 0.86)
- Deep RL Needs Deep Behavior Analysis: Exploring Implicit Planning by Model-Free Agents in Open-Ended Environments (NeurIPS 2025 · similarity 0.85)
- ALINE: Joint Amortization for Bayesian Inference and Active Data Acquisition (NeurIPS 2025 · similarity 0.85)
- ACTIVE-o3 : Empowering MLLMs with Active Perception via Pure Reinforcement Learning (ICML 2026 · similarity 0.85)
- Breaking the Performance Ceiling in Reinforcement Learning requires Inference Strategies (NeurIPS 2025 · 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.