{"id":"W4407590589","doi":"10.48550/arxiv.2502.09589","title":"Logical forms complement probability in understanding language model (and human) performance","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Complement (music); Computer science; Logical conjunction; Natural language processing; Programming language; Chemistry","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01053266,0.0009017027,0.0004861417,0.002155268,0.0005764958,0.005095913,0.0009767156,0.001330335,0.004742728],"category_scores_gemma":[0.09527332,0.0005808253,0.0008801331,0.001455156,0.002277387,0.01121246,0.002603745,0.002218151,0.00123185],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001011821,"about_ca_system_score_gemma":0.0009044782,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004609214,"about_ca_topic_score_gemma":0.004409599,"domain_scores_codex":[0.9922414,0.004593046,0.0004359899,0.001556778,0.0009761191,0.0001967433],"domain_scores_gemma":[0.8829156,0.09794862,0.005933649,0.009970981,0.002099101,0.001131964],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.002147448,0.001139066,0.3505294,0.001851869,0.001102888,0.0004425747,0.01306992,0.09607648,0.01713734,0.06623803,0.01158586,0.4386792],"study_design_scores_gemma":[0.0001582789,0.0009525608,0.1276411,0.0003037563,0.0002838339,0.0007406964,0.003671742,0.5478312,0.01104918,0.287563,0.01951189,0.0002927627],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7431851,0.002085442,0.2280222,0.002748547,0.0001008194,0.0002056509,0.003558177,0.001918967,0.01817505],"genre_scores_gemma":[0.9475165,0.0003789115,0.04805465,0.0002386483,0.00004676914,0.00008602226,0.002964741,0.0001863911,0.0005273308],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01053266,"threshold_uncertainty_score":0.05570269,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1211045059254114,"score_gpt":0.337995929592267,"score_spread":0.2168914236668557,"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."}}