arXiv cs.AI· Muhammad Ahmed Mohsin, Saad Bin Ather, Valentin Six, Lancelot Da Costa, Tommaso Salvatori, Bernhard Scholkopf·· 1 小时前AI 评分30
语义先验何时进入部分可识别的因果发现？LLM 辅助框架研究
When Should Semantic Priors Enter Partially-Identified Causal Discovery?
AI 导读
一项 arXiv 研究提出框架，考察在部分可识别的因果发现中如何分配 LLM 语义查询、准入提议及保留因果图。基准网络实验显示，语义信息优于随机翻转方向，但针对不稳定边的定向分配相较随机分配无明确优势；所测准入证书无法支持非平凡准入，且学习到的参考仅在 80 个计划任务中的 3 个包含真实 DAG。
正文 · 原文
Abstract:Observational causal discovery often leaves causal di- rections unresolved or unstable. Large language mod- els (LLMs) can provide semantic information to resolve these directions, but incorrect proposals can exclude the true causal graph and invalidate downstream inference. We study three separate decisions: where to allocate semantic queries, which proposals to admit, and which causal graphs to retain. Our framework distinguishes population non-identifiability from finite-sample insta- bility, evaluates semantic orientation against statistical and random controls, and examines calibrated admis- sion and error-tolerant DAG retention. Experiments across benchmark networks show that semantic infor- mation improves recovery over coin-flip orientations, but targeting unstable edges has no established ad- vantage over matched-random allocation. The tested admission certificate cannot support nontrivial admis- sion with the available calibration tasks. Moreover, downstream retention depends on reference adequacy: learned references contain the true DAG in only three of 80 planned tasks. These findings separate the value of semantic information from the reliability of its alloca- tion, admission, and retention, and identify calibration size and reference adequacy as key limitations.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.14736 [cs.AI] |
| (or arXiv:2611.14736v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.14736 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Muhammad Ahmed Mohsin [view email]
[v1]
Tue, 10 Nov 2026 20:55:34 UTC (196 KB)
来源：arXiv cs.AI · arxiv.org