{"id":"W4402705760","doi":"10.48550/arxiv.2408.15417","title":"Implicit Geometry of Next-token Prediction: From Language Sparsity Patterns to Model Representations","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Security token; Computer science; Geometry; Artificial intelligence; Mathematics; Computer network","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":[0.00192057,0.000776969,0.0009344834,0.0007831659,0.0006226844,0.001592914,0.001583033,0.001014981,0.001834233],"category_scores_gemma":[0.01889496,0.0007773483,0.0006919443,0.0009648376,0.002014573,0.004685735,0.002797603,0.003316194,0.000711704],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001060578,"about_ca_system_score_gemma":0.001073512,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004273897,"about_ca_topic_score_gemma":0.006546946,"domain_scores_codex":[0.9988356,0.0005952823,0.00004942727,0.0002830823,0.0001332725,0.0001033481],"domain_scores_gemma":[0.9920322,0.005340686,0.0007179772,0.001148786,0.0004675887,0.0002927941],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003824253,0.0002236188,0.01601847,0.0002903933,0.0001360313,0.0003641266,0.0008515051,0.6765315,0.01001048,0.06781907,0.007197574,0.2201748],"study_design_scores_gemma":[0.000006529819,0.00002194121,0.0004411057,0.000009776167,0.000004146526,0.00002392087,0.00003976695,0.9536313,0.000867659,0.04458082,0.0003646006,0.000008547295],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.09582778,0.0003016422,0.9002243,0.001177101,0.00003912106,0.00004789871,0.0003527331,0.0009150501,0.0011143],"genre_scores_gemma":[0.8250816,0.0003170602,0.1706634,0.0003237385,0.00009914026,0.0001405442,0.001418174,0.0002645412,0.00169177],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004273897,"threshold_uncertainty_score":0.01015705,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0584834549110922,"score_gpt":0.2321547135932132,"score_spread":0.173671258682121,"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."}}