{"id":"W4385571325","doi":"10.18653/v1/2023.acl-long.230","title":"A fine-grained comparison of pragmatic language understanding in humans and language models","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":47,"is_retracted":false,"has_abstract":true,"ca_institutions":"Carleton University","funders":"","keywords":"Literal (mathematical logic); Computer science; Pragmatics; Heuristic; Set (abstract data type); Language model; Natural language processing; Artificial intelligence; Human language; Linguistics","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":[0.006100692,0.0006105062,0.0006064854,0.0008521196,0.0004932799,0.003236783,0.0007753965,0.00130573,0.001587585],"category_scores_gemma":[0.04243857,0.0005238039,0.0004884874,0.0003485536,0.002050514,0.004621082,0.002565439,0.001258862,0.000394052],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006612149,"about_ca_system_score_gemma":0.0007427527,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001761899,"about_ca_topic_score_gemma":0.002104661,"domain_scores_codex":[0.9939778,0.003379643,0.0002657149,0.001316868,0.0008540089,0.0002060998],"domain_scores_gemma":[0.9701812,0.02220559,0.001678066,0.004268933,0.001127704,0.0005384616],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"observational","study_design_scores_codex":[0.002491828,0.001526126,0.2352244,0.001823319,0.0008628598,0.0009919369,0.08378156,0.1000201,0.282074,0.04305609,0.005136681,0.243011],"study_design_scores_gemma":[0.0001902295,0.002503868,0.207912,0.0002350478,0.0002157436,0.001354512,0.02273107,0.5935327,0.06742853,0.09088837,0.0124764,0.0005315287],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9384727,0.0001370469,0.05655893,0.0002963755,0.00001962105,0.00008832115,0.0002410823,0.0004347048,0.003751246],"genre_scores_gemma":[0.9854631,0.00004657582,0.01380777,0.00006119257,0.000004723289,0.00004529221,0.0002651137,0.00005242676,0.0002538634],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.006100692,"threshold_uncertainty_score":0.03226399,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07989764519677654,"score_gpt":0.3256151878462655,"score_spread":0.2457175426494889,"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."}}