{"id":"W4398795962","doi":"10.48550/arxiv.2405.13718","title":"Next-token prediction capacity: general upper bounds and a lower bound for transformers","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Advancements in Photolithography Techniques","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Security token; Computer science; Transformer; Computer network; Electrical engineering; Engineering; Voltage","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":[0.0001217862,0.000399302,0.0003139988,0.0003932365,0.0001148715,0.0001299902,0.0002357584,0.0003834783,0.00002709402],"category_scores_gemma":[0.000004718215,0.0004845593,0.0002588574,0.0002598362,0.0001838331,0.0002788194,0.0001757041,0.0006078124,0.000003903609],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002918889,"about_ca_system_score_gemma":0.00003824594,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000465397,"about_ca_topic_score_gemma":0.00002914477,"domain_scores_codex":[0.9986427,0.00001808021,0.0002129636,0.0006657196,0.00007144728,0.0003890242],"domain_scores_gemma":[0.9993938,0.00002863821,0.00004253317,0.0003488311,0.00005725067,0.0001289272],"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":[0.001378201,0.0007009174,0.01461366,0.019717,0.01045312,0.0007634852,0.00532794,0.5238957,0.01897991,0.3498365,0.02607295,0.02826059],"study_design_scores_gemma":[0.001003537,0.0002340598,0.0004381651,0.0004943057,0.0009355699,0.0000191365,0.0001845364,0.6773476,0.002786467,0.2666082,0.04859018,0.001358223],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7670868,0.0004682073,0.2236341,0.00001519235,0.001419683,0.0009183893,0.0005271865,0.001205816,0.004724655],"genre_scores_gemma":[0.9943905,0.001547195,0.002983944,0.00003034279,0.0001580972,0.00002616528,0.00007290643,0.00009359806,0.0006972366],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2273037,"threshold_uncertainty_score":0.9997606,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04204735105369126,"score_gpt":0.1918230236967434,"score_spread":0.1497756726430522,"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."}}