{"id":"W6891814438","doi":"10.48448/h82z-es28","title":"Clever Hans or Neural Theory of Mind? Stress Testing Social Reasoning in Large Language Models","year":2024,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Heuristics; Adversarial system; Artificial neural network; Benchmark (surveying); Work (physics); Stress (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":[0.007423249,0.0006324571,0.0003300124,0.0005733299,0.0007324545,0.002181415,0.001247621,0.001033063,0.006768043],"category_scores_gemma":[0.06586231,0.0001853489,0.0003666588,0.000500483,0.002437493,0.003065449,0.001689409,0.001900406,0.00104441],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00101645,"about_ca_system_score_gemma":0.001067037,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002894063,"about_ca_topic_score_gemma":0.005280139,"domain_scores_codex":[0.9954026,0.003210948,0.0001236789,0.0004510976,0.0007046169,0.0001071483],"domain_scores_gemma":[0.9553971,0.0333719,0.002216744,0.006393589,0.001669852,0.0009507889],"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":[0.001149149,0.001864152,0.1013715,0.001013557,0.0005574777,0.0008754995,0.005325755,0.1984428,0.01104763,0.1702611,0.04799714,0.4600942],"study_design_scores_gemma":[0.00010201,0.0005762087,0.01677768,0.0002374392,0.00008670391,0.000272333,0.001892671,0.6425365,0.01129696,0.3074809,0.01866359,0.00007704225],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7909825,0.001116011,0.1044141,0.01695112,0.0002419582,0.0001697408,0.001149982,0.001920537,0.08305413],"genre_scores_gemma":[0.9723632,0.0001605884,0.02377792,0.0005751053,0.00004605878,0.00007553297,0.0005601011,0.0001176346,0.002323875],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.007423249,"threshold_uncertainty_score":0.03925836,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05831946151645435,"score_gpt":0.3393438508082843,"score_spread":0.28102438929183,"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."}}