{"id":"W3136363192","doi":"10.1016/j.csl.2022.101429","title":"On the effect of dropping layers of pre-trained transformer models","year":2022,"lang":"en","type":"article","venue":"Computer Speech & Language","topic":"Topic Modeling","field":"Computer Science","cited_by":91,"is_retracted":false,"has_abstract":false,"ca_institutions":"Dalhousie University","funders":"","keywords":"Transformer; Computer science; Limiting; Sentence; Artificial intelligence; Paraphrase; Machine learning; Natural language processing; Engineering","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0006593522,0.0001446029,0.0002542182,0.0001089619,0.0001013203,0.00002545134,0.001276457,0.00002805312,0.00006145281],"category_scores_gemma":[0.0000102901,0.0001032252,0.0001501285,0.0002885147,0.00003259606,0.000136569,0.0002630788,0.0002241165,0.000001923246],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003700953,"about_ca_system_score_gemma":0.00002978869,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006363192,"about_ca_topic_score_gemma":0.000002771019,"domain_scores_codex":[0.9985176,0.0002245439,0.0002811317,0.0002972712,0.0004559568,0.000223472],"domain_scores_gemma":[0.9987996,0.0002898316,0.0001141221,0.0007357628,0.00002209999,0.00003863245],"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":[0.0001081229,0.0001395528,0.00008455669,0.0001957385,0.0001485282,0.000101798,0.03234697,0.2837707,0.01595713,0.0798124,0.0006810533,0.5866534],"study_design_scores_gemma":[0.0005144819,0.0005873109,0.00008251704,0.00003212707,0.00001028042,0.00002077758,0.00006993425,0.9647587,0.03249719,0.001220454,0.00006964427,0.000136568],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4287985,0.00007406495,0.5692059,0.0002471228,0.0002038043,0.0002393578,0.000004269283,0.00005376849,0.001173182],"genre_scores_gemma":[0.9818287,0.00000118989,0.01775303,0.0002619892,0.00004996991,0.00002064413,0.00000200001,0.00001129785,0.00007116513],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.680988,"threshold_uncertainty_score":0.4209401,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01020937020698539,"score_gpt":0.2361752071494904,"score_spread":0.225965836942505,"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."}}