{"id":"W4389519946","doi":"10.18653/v1/2023.findings-emnlp.300","title":"Verb Conjugation in Transformers Is Determined by Linear Encodings of Subject Number","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"York University; National Science Foundation","keywords":"Verb; Subject (documents); Computer science; Transformer; Encoding (memory); Mathematics Subject Classification; Position (finance); Artificial intelligence; Layer (electronics); Linguistics; Natural language processing; Arithmetic; Speech recognition; Mathematics; Discrete mathematics; Philosophy; Physics","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":[],"consensus_categories":[],"category_scores_codex":[0.0002019299,0.00005861444,0.00009686039,0.00006758417,0.00001591357,0.00001388432,0.0002278142,0.00003960108,0.00005977162],"category_scores_gemma":[0.00001339975,0.00005559875,0.00003239478,0.000409591,0.00001302245,0.0002512319,0.00003019251,0.00004877934,0.00007417867],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001842211,"about_ca_system_score_gemma":0.00002737501,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002529886,"about_ca_topic_score_gemma":0.00002786821,"domain_scores_codex":[0.999297,0.00001438508,0.0002020946,0.0001836615,0.0001537709,0.0001490455],"domain_scores_gemma":[0.9997252,0.00005102915,0.0000290545,0.0001440188,0.00002357014,0.00002715812],"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":[0.000079933,0.0002679801,0.1402026,0.0004350571,0.00008017628,0.00006023572,0.05795074,0.001956735,0.2505713,0.03800791,0.022564,0.4878233],"study_design_scores_gemma":[0.0005109866,0.00002767122,0.001121619,0.00002091062,0.00000188889,0.000001841068,0.00008569592,0.8843151,0.1118023,0.0009692807,0.001014547,0.0001281587],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6987279,0.000003512225,0.2956189,0.0006643272,0.00009206666,0.00008681484,0.000001458096,0.00009509593,0.004709979],"genre_scores_gemma":[0.9912466,0.00001355566,0.007696581,0.000203567,0.000009070108,0.000004655257,0.000002046784,0.000003813083,0.000820112],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8823584,"threshold_uncertainty_score":0.2267251,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01973994559083426,"score_gpt":0.269174178691716,"score_spread":0.2494342331008818,"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."}}