Research map / Probabilistic and scientific modeling / Causal discovery
Test-Time Learning of Causal Structure from Interventional Data
Wei Chen, Rui Ding, Huang Bojun, Yang Zhang, Qiang Fu, Yuxuan Liang, Shi Han, Dongmei Zhang
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 |
|---|---|---|---|
| Generate training data with IS-MCMC self-augmentation instead of purely random graphs (ablation) | TICL · bnlearn Asia, Sachs | SHD, SID, F1 | Improved |
| Initialize IS-MCMC with proxy-algorithm seed graph instead of random seed (ablation) | TICL IS-MCMC · bnlearn Earthquake, Sachs | convergence, F1, SHD, SID | Improved |
| Increase number of training instances (100 to 1600) (ablation) | TICL · bnlearn Earthquake, Survey, Asia, Sachs | SHD, SID, F1 | Improved |
| Increase test sample size per instance (1k to 20k) (ablation) | TICL · bnlearn Earthquake, Survey, Asia, Sachs | SHD, SID, F1 | Improved |
| Use multi-chain parallel MCMC (ablation) | IS-MCMC · sampling time vs graph size | sampling time | Improved |
| Apply TICL (test-time training + JCI + two-phase SCL) versus best baselines | TICL · bnlearn intervention target detection | F1: average F1 improvement of 50.21% over second-best | Improved |
| Apply TICL versus best baseline on I-CPDAG discovery | TICL · bnlearn I-CPDAG | F1: 13.62% over the best baseline | Improved |
| Wrap observational methods in JCI framework | PC, HC, BLIP, GOLEM · bnlearn | F1 rank | Improved |
| Use TICL on Pathfinder versus ENCO with known targets | TICL · bnlearn Pathfinder | SHD | No clear effect |
Closest papers by meaning
- Efficient Causal Structure Learning from Multiple Interventional Datasets with Unknown Targets (AAAI 2022 · similarity 0.89)
- CausalTab: Pretraining Across Causal Environments for Tabular Causal Discovery (NeurIPS 2026 · similarity 0.88)
- Causal Discovery from Soft Interventions with Unknown Targets: Characterization and Learning (NeurIPS 2020 · similarity 0.88)
- Nonparametric Identifiability of Causal Representations from Unknown Interventions (NeurIPS 2023 · similarity 0.87)
- Towards Causal Foundation Model: on Duality between Optimal Balancing and Attention (ICML 2024 · similarity 0.86)
- Learning Mixtures of Unknown Causal Interventions (NeurIPS 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.
