Research map / Reinforcement learning / Visual and representation RL
Distributional Active Inference
Abdullah Akgül, Gulcin Baykal, Manuel Haussmann, Mustafa Mert Çelikok, Melih Kandemir
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add the active inference component to DTD3 (DTD3 -> DAIF) (ablation) | DTD3 distributional actor-critic · EvoGym | AULC rank: 2.0 -> 1.5 | Improved |
| add the active inference component to DTD3 (DTD3 -> DAIF) (ablation) | DTD3 distributional actor-critic · EvoGym | Final Return rank: 2.0 -> 1.6 | Improved |
| add the active inference component to DTD3 (DTD3 -> DAIF) (ablation) | DTD3 distributional actor-critic · DMC | AULC rank: 2.4 -> 1.6 | Improved |
| add the active inference component to DTD3 (DTD3 -> DAIF) (ablation) | DTD3 distributional actor-critic · DMC | Final Return rank: 2.6 -> 1.5 | Improved |
| add the active inference component to DTD3 (DTD3 -> DAIF) (ablation) | DTD3 distributional actor-critic · DMC Vision | AULC rank: 2.9 -> 1.9 | Improved |
| add the active inference component to DTD3 (DTD3 -> DAIF) (ablation) | DTD3 distributional actor-critic · DMC Vision | Final Return rank: 3.8 -> 2.0 | Improved |
| add SAC-style entropy regularization to DAIF (ablation) | DAIF · continuous control | return | Hurt |
| use DAIF instead of DTD3 in wall-clock compute | DAIF · continuous control | wall-clock overhead: about 12% more wall-clock computation time | Hurt |
| use DSAC or DRND instead of DTD3 (compute overhead) | DTD3 · continuous control | wall-clock overhead: DSAC (26%) and DRND (37%) relative to DTD3 | Hurt |
Closest papers by meaning
- Contrastive Active Inference (NeurIPS 2021 · similarity 0.89)
- Deep active inference agents using Monte-Carlo methods (NeurIPS 2020 · similarity 0.89)
- Making Sense of Reinforcement Learning and Probabilistic Inference (ICLR 2020 · similarity 0.87)
- ACTIVE-o3 : Empowering MLLMs with Active Perception via Pure Reinforcement Learning (ICML 2026 · similarity 0.87)
- Breaking the Performance Ceiling in Reinforcement Learning requires Inference Strategies (NeurIPS 2025 · similarity 0.86)
- Geometric Active Exploration in Markov Decision Processes: the Benefit of Abstraction (ICML 2024 · 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.
