{"id":"W4415981209","doi":"10.48550/arxiv.2505.15962","title":"Pre-training Limited Memory Language Models with Internal and External Knowledge","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Division of Materials Research; Materials Research Science and Engineering Center, Harvard University; National Science Foundation; Advanced Research Projects Agency; Natural Sciences and Engineering Research Council of Canada; Cornell Center for Materials Research; Defense Advanced Research Projects Agency","keywords":"Memorization; Encoding (memory); Class (philosophy); Language model; Memory model; Knowledge acquisition; Knowledge representation and reasoning","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0003223351,0.000351739,0.0003960803,0.0002181613,0.0000904156,0.0002125342,0.001419611,0.0002055832,0.000008796363],"category_scores_gemma":[0.00003325836,0.0003122346,0.00008418269,0.0001413656,0.00006557373,0.0003300504,0.002881384,0.0008505619,0.000009101826],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007647035,"about_ca_system_score_gemma":0.0003004021,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002544413,"about_ca_topic_score_gemma":0.00009194715,"domain_scores_codex":[0.9979376,0.0000974394,0.0003546285,0.0009892023,0.0002414241,0.0003797472],"domain_scores_gemma":[0.9983847,0.0001074787,0.0001759916,0.001085434,0.0001048316,0.0001415343],"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":[0.000187344,0.0004034916,0.1424214,0.001900711,0.0008262185,0.0008432148,0.2267449,0.08701213,0.002483125,0.01887829,0.0003136268,0.5179855],"study_design_scores_gemma":[0.0006566402,0.00006111291,0.0158722,0.002040187,0.00004976166,0.00006282262,0.0003263797,0.9766248,0.0009196565,0.002708975,0.00008505186,0.0005923879],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5113002,0.001151502,0.4825085,0.00009936545,0.0004161884,0.0001566378,0.000003831863,0.0001809898,0.004182786],"genre_scores_gemma":[0.9414883,0.00004825638,0.05414728,0.0002278771,0.0002725787,0.0000421628,0.000003924208,0.00001881979,0.003750822],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8896127,"threshold_uncertainty_score":0.9999329,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05873123808312963,"score_gpt":0.2915046928211787,"score_spread":0.2327734547380491,"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."}}