{"id":"W4409963278","doi":"10.1098/rsos.241776","title":"Generalization bias in large language model summarization of scientific research","year":2025,"lang":"en","type":"article","venue":"Royal Society Open Science","topic":"Topic Modeling","field":"Computer Science","cited_by":84,"is_retracted":false,"has_abstract":true,"ca_institutions":"Western University","funders":"","keywords":"Automatic summarization; Generalization; Computer science; Natural language processing; Artificial intelligence; Language model; Epistemology; Philosophy","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.009129119,0.00007311653,0.0001281775,0.0002422142,0.0006123139,0.0009666612,0.004083626,0.00005457171,0.000007267462],"category_scores_gemma":[0.0003564683,0.00007094902,0.00004071261,0.004883278,0.0003604509,0.001012872,0.002953598,0.0001558515,0.000004767589],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001964008,"about_ca_system_score_gemma":0.001146467,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005092187,"about_ca_topic_score_gemma":0.0001810521,"domain_scores_codex":[0.9975981,0.0001042314,0.0002807285,0.0006884224,0.0008871639,0.0004414251],"domain_scores_gemma":[0.9985739,0.00006571107,0.00006569149,0.0007993391,0.0004398441,0.00005554334],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000003700737,0.0002100611,0.02222649,0.00005725117,0.000004688839,0.000001323699,0.01638323,0.2637357,0.02766241,0.6556093,0.00252709,0.01157878],"study_design_scores_gemma":[0.0002115361,0.000006311882,0.002257053,0.00004179902,7.198365e-7,7.801214e-8,0.0003190004,0.9865518,0.006219788,0.004215802,0.0001086653,0.00006751046],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2533852,0.00005635896,0.7406198,0.000440178,0.0001770127,0.0002969101,0.000002594829,0.00002273635,0.004999209],"genre_scores_gemma":[0.9181766,0.000004443775,0.07779441,0.0001509233,0.000007742886,0.00001479031,0.000003645861,0.000002852055,0.003844609],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.722816,"threshold_uncertainty_score":0.9321536,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09027618539005793,"score_gpt":0.3897188774648272,"score_spread":0.2994426920747693,"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."}}