{"id":"W2915581179","doi":"10.48550/arxiv.1902.07249","title":"Discovery of Natural Language Concepts in Individual Units of CNNs","year":2019,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Topic Modeling","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"","keywords":"Computer science; Morpheme; Natural language processing; Artificial intelligence; Natural language; Natural (archaeology); Translation (biology); Machine translation; Linguistics; History","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":[0.0002036789,0.0001790636,0.0003732865,0.0003107806,0.00001458513,0.00002813058,0.001799185,0.0001827803,0.000004687693],"category_scores_gemma":[0.00004621102,0.000200813,0.0001001528,0.000569021,0.00008929445,0.0004226979,0.002177359,0.0004739184,0.000004980813],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007539869,"about_ca_system_score_gemma":0.0002967191,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000542098,"about_ca_topic_score_gemma":0.0001088219,"domain_scores_codex":[0.9987252,0.0001148984,0.0002470801,0.000579809,0.000122794,0.0002102102],"domain_scores_gemma":[0.9984471,0.0001201718,0.0003142817,0.0009797759,0.0001040381,0.00003461044],"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":[0.00005483846,0.0001862501,0.06998217,0.0006027194,0.0001968968,0.0003288572,0.01061381,0.6094332,0.001058104,0.3051521,0.00007013096,0.002320903],"study_design_scores_gemma":[0.001835281,0.00009370168,0.02566415,0.0007816673,0.00007824441,0.000004175486,0.00180608,0.9551558,0.005910908,0.007844657,0.00003499071,0.0007903335],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8731672,0.0001826727,0.1251639,0.00001353787,0.0004714277,0.0001710875,0.00003391243,0.00002938973,0.0007668677],"genre_scores_gemma":[0.9981851,0.00002581086,0.000948192,0.0000223338,0.00002129874,1.812997e-7,0.0000160398,0.000007585043,0.0007734872],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3457226,"threshold_uncertainty_score":0.8188916,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07753615855421307,"score_gpt":0.2152653111743964,"score_spread":0.1377291526201834,"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."}}