Research map / Probabilistic and scientific modeling / Causal discovery
Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy to Game
Alexander Gilbert Reisach, Christof Seiler, Sebastian Weichwald
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 |
|---|---|---|---|
| standardize the data (removing variance scale information) | NOTEARS, MSE-GDS, GOLEM-EV · simulated linear ANMs, ER-2 Gaussian-NV etc. | SID/SHD | Hurt |
| standardize the data | GOLEM-NV · simulated linear ANMs | SID/SHD | Improved |
| standardize the data | PC, FGES, DirectLiNGAM · simulated linear ANMs | SID/SHD | No clear effect |
| replace continuous structure learners with sortnregress (sort by variance then lasso) | sortnregress · simulated raw ANMs with high varsortability | SID/SHD | No clear effect |
| sort by variance instead of random order | variance sorting · non-linear GP ANMs ER-1 | SID: 27.9 -> 7.7 (lower SID better) | Improved |
| sort by variance vs random on dense non-linear graph ER-4 | variance sorting · non-linear GP ANMs ER-4 | SID: 63.1 -> 25.2 | Improved |
| sort by variance vs random on SF-4 | variance sorting · non-linear GP ANMs SF-4 | SID: 59.5 -> 7.6 | Improved |
| apply varsortability-exploiting methods to real data (varsortability near chance) | sortnregress / continuous methods · protein signaling dataset | performance pattern | No clear effect |
| chain orientation rule using increasing regression coefficients | decision rule · standardized and scale-harmonized chains | orientation accuracy | Improved |
Closest papers by meaning
- Boosting Causal Structure Learning: An Asymmetric Exponential Modulation Gaussian-Based Adaptive Sample Reweighting Framework (AAAI 2025 · similarity 0.89)
- Boosting Causal Discovery via Adaptive Sample Reweighting (ICLR 2023 · similarity 0.89)
- Effective Causal Discovery under Identifiable Heteroscedastic Noise Model (AAAI 2024 · similarity 0.88)
- CoLiDE: Concomitant Linear DAG Estimation (ICLR 2024 · similarity 0.87)
- Standardizing Structural Causal Models (ICLR 2025 · similarity 0.87)
- Strong and Weak Identifiability of Optimization-based Causal Discovery in Non-linear Additive Noise Models (ICML 2025 · 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.
