{"id":"W4414359425","doi":"10.24963/ijcai.2025/1186","title":"Large Language Models for Causal Discovery: Current Landscape and Future Directions","year":2025,"lang":"en","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Transformative learning; Causality (physics); Leverage (statistics); Key (lock); Natural language; Metadata; Causal model","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.06840863,0.001743624,0.003960904,0.006560255,0.001726157,0.01156869,0.006267826,0.005524188,0.01186505],"category_scores_gemma":[0.1635459,0.002092304,0.003064452,0.007625748,0.009044569,0.02948508,0.007096949,0.01199919,0.003122105],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.006153386,"about_ca_system_score_gemma":0.01039756,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01134632,"about_ca_topic_score_gemma":0.009821074,"domain_scores_codex":[0.972966,0.02099177,0.0008697525,0.002212659,0.002555843,0.0004039924],"domain_scores_gemma":[0.526693,0.4499796,0.003293498,0.0101379,0.008087778,0.001808133],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"not_applicable","study_design_scores_codex":[0.000247021,0.0002232969,0.00549556,0.004539087,0.0004340578,0.0001348646,0.001001269,0.0223574,0.0002018813,0.5375536,0.02133291,0.406479],"study_design_scores_gemma":[0.00005400165,0.00006055856,0.0006550028,0.00218229,0.00009985676,0.0000982209,0.0006357087,0.06290155,0.0001420338,0.8788,0.05429313,0.00007770395],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"review","genre_scores_codex":[0.005308195,0.3700281,0.4523562,0.1573994,0.0007921136,0.0002228787,0.001141972,0.001795031,0.01095599],"genre_scores_gemma":[0.1295679,0.3753352,0.4672419,0.01458821,0.006719369,0.001329414,0.002015702,0.0006459708,0.002556347],"genre_candidate":"review","genre_consensus":null,"teacher_disagreement_score":0.06840863,"threshold_uncertainty_score":0.3617837,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01076595841234547,"score_gpt":0.2922617763234516,"score_spread":0.2814958179111062,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}