Research map / Probabilistic and scientific modeling / Causal discovery
Systematic Evaluation of Causal Discovery in Visual Model Based 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 |
|---|---|---|---|
| increase number of objects from 3 to 5 (ablation) | Modular model · Physics, 10-step prediction | ranking metric: 90 -> 50 | Hurt |
| use explicit structure (GNN/modular) instead of monolithic models | monolithic AE/VAE · Physics and Chemistry | prediction metrics and RL success | Helped |
| scale from 3 to 5 objects for GNNs (ablation) | GNN · Physics | prediction metrics and RL performance | Hurt |
| train with contrastive loss instead of likelihood loss | structured models · Physics and Chemistry | ranking metric | Helped |
| train with contrastive loss instead of likelihood loss | world models · Physics and Chemistry | downstream RL performance | Mixed |
| use modular models instead of other models on chain and full graphs | Modular model · Chemistry | prediction metrics and RL | Helped |
| use unobserved variable setting instead of observed (ablation) | all models · Physics downstream RL | success rate | Hurt |
| use ranking metrics as proxy for downstream RL performance | AE with NLL Finetune · Physics | downstream RL | No clear effect |
Closest papers by meaning
- Learning Linear Causal Representations from General Environments: Identifiability and Intrinsic Ambiguity (NeurIPS 2024 · similarity 0.90)
- CausalVerse: Benchmarking Causal Representation Learning with Configurable High-Fidelity Simulations (NeurIPS 2025 · similarity 0.90)
- Causal Representation Learning Made Identifiable by Grouping of Observational Variables (ICML 2024 · similarity 0.89)
- Neural Causal Abstractions (AAAI 2024 · similarity 0.89)
- Hierarchical Reinforcement Learning with Targeted Causal Interventions (ICML 2025 · similarity 0.89)
- A Fixed-Point Approach for Causal Generative Modeling (ICML 2024 · similarity 0.89)
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.