LLMs can draw DAGs but can't guarantee d-separation, notebook discussion
fdellaert · x · 2026-09-21
A technical exchange on LLMs doing Bayesian network structure learning. @jatingargiitk notes an LLM can specify the structure if given examples and variable names, but the harder part is making it respect conditional independence without violating d-separation rules.
Frank Dellaert replies that the notebook performs a search over DAGs, so the distribution is factorized over the selected graph with d-separation holding by construction — but that does not mean it finds the right graph. The search guarantees graph-decomposition consistency, not correctness of the causal structure.
Related event: LLMs can draft DAGs but can't guarantee d-separation(2 posts)→
More from Research
- GPT-6 Astra claims Terminal-Bench Science lead at 65.7%, 31.4 points clear of second place — DeryaTR_ · 2026-09-21
- Ant International's Code2Skill mines verifiable agent skills from code at scale — ant-intl · 2026-09-21
- RecreationWorld: five-platform CUA benchmark — GPT-6 Astra hits 58.1% but passes all tests on just 2.8% — Shuai Bai · 2026-09-21
- OmniVBench: a 12k-checklist benchmark and 340K-sample dataset for omni reference-to-video generation — Wenxue Li · 2026-09-21
- AI reviews training AI reviewers: study finds 'scientific-judgment collapse' and an open-source fix — Sy-Tuyen Ho · 2026-09-21
- Apple's MintAct unifies GUI agents across mobile, desktop, and web, hitting SOTA 48.9 on OSWorld-Verified — apple · 2026-09-21