{"id":"W6929356424","doi":"10.48448/mrtp-zn19","title":"Recursive Neural Networks with Bottlenecks Diagnose (Non-)Compositionality","year":2022,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Nonlinear Optical Materials Studies","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Principle of compositionality; Bottleneck; Information bottleneck method; Metric (unit); Artificial neural network; Ranking (information retrieval); Natural language","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":["metaepi_narrow","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.000234605,0.0003315932,0.000362204,0.000211012,0.0002258365,0.00009881458,0.0005342882,0.00009833861,0.002428859],"category_scores_gemma":[0.00003490505,0.0002821578,0.0000398168,0.0005282314,0.0007402918,0.000097836,0.0002765698,0.0003696759,0.00004251125],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000220676,"about_ca_system_score_gemma":0.00007579447,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006992688,"about_ca_topic_score_gemma":0.0001476722,"domain_scores_codex":[0.9982308,0.00002743559,0.0002198001,0.0004616706,0.0005406642,0.0005196062],"domain_scores_gemma":[0.9992266,0.00009456032,0.00007503532,0.0004066524,0.00005778846,0.0001393599],"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":[0.00008245851,0.0004639831,0.00145152,0.000500601,0.0005152567,0.0004021297,0.0003266354,0.5330653,0.001056565,0.006014512,0.4477491,0.008371962],"study_design_scores_gemma":[0.001890308,0.0009469271,0.00250869,0.0005528037,0.0003090065,0.0001480034,0.000447992,0.6230481,0.0005459673,0.001185232,0.3647544,0.003662532],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.01169002,0.004640304,0.02493503,0.00132951,0.009846099,0.002565723,0.001078948,0.003882684,0.9400317],"genre_scores_gemma":[0.6540535,0.003207092,0.2009599,0.003186051,0.01558202,0.001520278,0.00246468,0.005395729,0.1136307],"genre_candidate":"other","genre_consensus":null,"teacher_disagreement_score":0.826401,"threshold_uncertainty_score":0.999963,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.006796463485454486,"score_gpt":0.2277335413085808,"score_spread":0.2209370778231263,"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."}}