{"id":"W4412945079","doi":"10.18653/v1/2025.acl-long.824","title":"Logical forms complement probability in understanding language model (and human) performance","year":2025,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Complement (music); Computer science; Programming language; Natural language processing; Artificial intelligence; Theoretical computer science","routes":{"ca_aff":true,"ca_fund":true,"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":[0.01000233,0.0009011496,0.0004679912,0.002215938,0.0005593224,0.004957847,0.0009556044,0.001265306,0.0044301],"category_scores_gemma":[0.09491963,0.0005612174,0.0008367563,0.0013427,0.002223976,0.01111031,0.002549417,0.001935772,0.001042917],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009286857,"about_ca_system_score_gemma":0.0008731089,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004783837,"about_ca_topic_score_gemma":0.004578358,"domain_scores_codex":[0.992768,0.004194521,0.0004476471,0.001462493,0.0009381651,0.0001891915],"domain_scores_gemma":[0.8805428,0.100108,0.006561498,0.009533357,0.002094225,0.001160079],"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":[0.002153933,0.001121065,0.4002649,0.001721947,0.001037438,0.0004373219,0.01332303,0.087147,0.01769894,0.05239472,0.008744162,0.4139555],"study_design_scores_gemma":[0.0001816137,0.001235143,0.1720627,0.0003308165,0.0003161807,0.0008872243,0.004540271,0.5570829,0.01256072,0.2305696,0.01990067,0.0003322098],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7825414,0.001733324,0.1924618,0.002141104,0.00008063022,0.0002224748,0.003006849,0.001684533,0.01612805],"genre_scores_gemma":[0.9527607,0.0003281472,0.04353718,0.0001972009,0.00003687335,0.00008200864,0.002461213,0.0001452728,0.0004513225],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01000233,"threshold_uncertainty_score":0.05289799,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05872521618947497,"score_gpt":0.3241288822758984,"score_spread":0.2654036660864234,"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."}}