![[2602.20333] DMCD: Semantic-Statistical Framework for Causal Discovery](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20333] DMCD: Semantic-Statistical Framework for Causal Discovery
Summary
The DMCD framework integrates LLM-based semantic drafting with statistical validation for causal discovery, enhancing performance across various real-world datasets.
Why It Matters
This research presents a novel approach to causal discovery that combines semantic reasoning with statistical methods, potentially improving data analysis in fields like industrial engineering and environmental monitoring. It highlights the importance of integrating machine learning with causal inference, a growing area of interest in AI.
Key Takeaways
- DMCD combines semantic drafting and statistical validation for causal discovery.
- The framework shows improved performance in recall and F1 scores across diverse datasets.
- Semantic reasoning over metadata is crucial for achieving better results.
- The approach is applicable to various fields, including engineering and IT systems.
- Phase I and II of DMCD enhance causal structure learning through a two-phase process.
Computer Science > Artificial Intelligence arXiv:2602.20333 (cs) [Submitted on 23 Feb 2026] Title:DMCD: Semantic-Statistical Framework for Causal Discovery Authors:Samarth KaPatel, Sofia Nikiforova, Giacinto Paolo Saggese, Paul Smith View a PDF of the paper titled DMCD: Semantic-Statistical Framework for Causal Discovery, by Samarth KaPatel and 3 other authors View PDF HTML (experimental) Abstract:We present DMCD (DataMap Causal Discovery), a two-phase causal discovery framework that integrates LLM-based semantic drafting from variable metadata with statistical validation on observational data. In Phase I, a large language model proposes a sparse draft DAG, serving as a semantically informed prior over the space of possible causal structures. In Phase II, this draft is audited and refined via conditional independence testing, with detected discrepancies guiding targeted edge revisions. We evaluate our approach on three metadata-rich real-world benchmarks spanning industrial engineering, environmental monitoring, and IT systems analysis. Across these datasets, DMCD achieves competitive or leading performance against diverse causal discovery baselines, with particularly large gains in recall and F1 score. Probing and ablation experiments suggest that these improvements arise from semantic reasoning over metadata rather than memorization of benchmark graphs. Overall, our results demonstrate that combining semantic priors with principled statistical verification yields a high-...