{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001640742,0.001371304,0.0009583234,0.0006015079,0.0004327428,0.002080557,0.003469927,0.001596404,0.004885897],"category_scores_gemma":[0.01294904,0.0008028093,0.001042875,0.0007247329,0.0008408425,0.007189348,0.002073614,0.003773412,0.002955547],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0010848,"about_ca_system_score_gemma":0.00155631,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004183177,"about_ca_topic_score_gemma":0.009360364,"domain_scores_codex":[0.9991844,0.0002289652,0.00006352195,0.0002966236,0.0001434761,0.00008291163],"domain_scores_gemma":[0.9941533,0.003639436,0.0002219403,0.00142177,0.0004105897,0.0001528646],"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":[0.0008728449,0.0005732047,0.005595439,0.0006286498,0.000331771,0.0004908304,0.0005152021,0.3906786,0.01862624,0.01506063,0.0200386,0.5465879],"study_design_scores_gemma":[0.00003858532,0.00009497989,0.0002591481,0.00003176289,0.000046497,0.00009063615,0.00005528358,0.9730334,0.007702481,0.01614887,0.002478367,0.00001998507],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2151628,0.002126445,0.7458196,0.002240931,0.0003652077,0.0002188029,0.00177696,0.02261139,0.009677839],"genre_scores_gemma":[0.7977693,0.0006244036,0.1886161,0.0007911628,0.000150593,0.0003417567,0.003721303,0.000750226,0.007235228],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004885897,"threshold_uncertainty_score":0.0163449,"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."}}