{"id":"W4389519826","doi":"10.18653/v1/2023.findings-emnlp.245","title":"FinePrompt: Unveiling the Role of Finetuned Inductive Bias on Compositional Reasoning in GPT-4","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"Institute for Information and Communications Technology Promotion; Ministry of Science and ICT, South Korea","keywords":"Computer science; Leverage (statistics); Task (project management); Inductive bias; Inductive reasoning; Question answering; Artificial intelligence; Human–computer interaction; Natural language processing; Multi-task learning","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004612308,0.001225695,0.0008565848,0.0006561324,0.0007367497,0.002486357,0.002384759,0.001686176,0.007066133],"category_scores_gemma":[0.02614509,0.0005835543,0.001153651,0.0005829036,0.001875257,0.006974253,0.004686947,0.003690864,0.002672669],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009830152,"about_ca_system_score_gemma":0.00201083,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00145344,"about_ca_topic_score_gemma":0.002669572,"domain_scores_codex":[0.9971479,0.001335772,0.0001583228,0.0006989015,0.0004954751,0.0001636783],"domain_scores_gemma":[0.9872362,0.008120116,0.0004212775,0.002840478,0.0009877111,0.0003941163],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.001624875,0.0006382288,0.006241924,0.002201474,0.0001526563,0.0009463835,0.007501113,0.09109602,0.0835075,0.1688365,0.02129038,0.615963],"study_design_scores_gemma":[0.0002317277,0.0003479182,0.0006623159,0.0001127146,0.00009290176,0.0003941314,0.0006338903,0.6905217,0.03653995,0.2437732,0.02661436,0.0000750761],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02895772,0.0002110986,0.9527691,0.0005941099,0.0001265622,0.0002447572,0.0004068609,0.01407803,0.002611775],"genre_scores_gemma":[0.3266837,0.0001800612,0.6660444,0.0005763509,0.0001094247,0.0003448489,0.001322377,0.001286433,0.003452425],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007066133,"threshold_uncertainty_score":0.02439255,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04108843627987746,"score_gpt":0.26247931949821,"score_spread":0.2213908832183325,"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."}}