Agent-guided Causal Discovery with a Small Language Model
Bibliographic record
Abstract
The task of inferring the causal structure of a system is essential for many applications, yet remains challenging for current methods. Recently, Large Language Models (LLMs) have been explored for causal discovery, but they are often impractical or infeasible to run locally. To address this, we propose a Small Language Model (SLM) agentic framework for causal discovery, consisting of cooperating agents. The first agent performs statistical tests between pairs of variables and, when a relationship is found, decides the causal direction. The next agent suggests candidate edges that are likely missing from the graph. Finally, a critic agent prunes spurious edges and outputs the final causal graph. We evaluate this framework on three benchmark causal graphs and show that it outperforms a one-shot SLM, "expert-in-loop" SLM, and purely data-driven methods such as PC and Greedy Equivalence Search (GES). These results demonstrate that an agentic SLM framework is effective for automated causal discovery, combining expert domain reasoning with statistical tools in a systematic and efficient way, while remaining practical to run on a personal computer with open-source local SLMs.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".