{"id":"W3207612418","doi":"10.18653/v1/2022.acl-short.24","title":"Kronecker Decomposition for GPT Compression","year":2022,"lang":"en","type":"preprint","venue":"","topic":"Algorithms and Data Compression","field":"Computer Science","cited_by":17,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Kronecker delta; Decomposition; Computer science; Kronecker product; Volume (thermodynamics); Compression (physics); Linguistics; Philosophy; Physics","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":[0.0002939248,0.0002691688,0.0003028786,0.0001398674,0.0003876298,0.0003301199,0.00199339,0.0001684656,0.0006818044],"category_scores_gemma":[0.00001168253,0.0002343507,0.0001912022,0.0001046746,0.00001883676,0.0003253615,0.007694897,0.0004548283,0.00002850544],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001047003,"about_ca_system_score_gemma":0.00011422,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00007234347,"about_ca_topic_score_gemma":0.000002549883,"domain_scores_codex":[0.9979306,0.0000894962,0.0003435191,0.0009221622,0.0004175126,0.0002967187],"domain_scores_gemma":[0.9980562,0.0001347435,0.0002164177,0.001390072,0.00009153596,0.0001110463],"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":[0.0001242721,0.0009708698,0.0001750098,0.0004771647,0.0001535489,0.00004141721,0.0008605466,0.02449653,0.002932996,0.1102518,0.312136,0.5473799],"study_design_scores_gemma":[0.0004506568,0.0001024576,0.0005299232,0.00009369567,0.00001497172,0.000007865918,0.00001475046,0.7990847,0.00115894,0.02081787,0.1772444,0.0004797758],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0009551317,0.0002818429,0.9900593,0.0007082595,0.002628257,0.0007235035,0.0001406953,0.0002910646,0.004211963],"genre_scores_gemma":[0.02882732,0.00006212663,0.9666107,0.0005843037,0.0003233426,0.0005516338,0.001462932,0.0000330654,0.001544583],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.7745882,"threshold_uncertainty_score":0.9591138,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02939809692835871,"score_gpt":0.3341453645357702,"score_spread":0.3047472676074115,"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."}}