{"id":"W4389363042","doi":"10.48550/arxiv.2312.00949","title":"Hyperparameter Optimization for Large Language Model Instruction-Tuning","year":2023,"lang":"en","type":"preprint","venue":"PolyPublie (École Polytechnique de Montréal)","topic":"Topic Modeling","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Mitacs; Hydro-Québec","keywords":"Hyperparameter; Computer science; Pipeline (software); Rank (graph theory); Set (abstract data type); Decomposition; Adaptation (eye); Machine learning; Language model; Artificial intelligence; Fine-tuning; Adaptive optimization; Programming language","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001026196,0.0005405526,0.0005735075,0.0007832999,0.0003356386,0.0006786794,0.001960074,0.0007722192,0.000007034801],"category_scores_gemma":[0.0003475407,0.0006055038,0.0003597859,0.0004463695,0.0000291266,0.0005622007,0.002279397,0.0009091822,0.0000102313],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005594776,"about_ca_system_score_gemma":0.0005045841,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001651518,"about_ca_topic_score_gemma":0.0005390853,"domain_scores_codex":[0.9964733,0.0001062994,0.0007314406,0.001253794,0.0004787508,0.0009564722],"domain_scores_gemma":[0.9966617,0.0001563562,0.0004396773,0.002263043,0.0002277362,0.0002515218],"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":[0.0000129469,0.00005278271,0.0002948274,0.0001166139,0.00004727757,0.00001298448,0.0007666904,0.9503304,0.0002331996,0.03839534,0.00060308,0.009133824],"study_design_scores_gemma":[0.000487247,0.00002956699,0.00008122392,0.0001280372,0.00003387525,0.00002225764,0.00006994953,0.9828498,0.0006316603,0.01495933,0.0001254379,0.0005816399],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01259081,0.0003681077,0.9791444,0.002445273,0.0008105618,0.001458542,0.0001599442,0.002916398,0.0001059812],"genre_scores_gemma":[0.1662721,0.00009380876,0.8296539,0.001016556,0.0002561406,0.001503611,0.0001613018,0.0001241358,0.0009183874],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.1536813,"threshold_uncertainty_score":0.9996396,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02511101953598357,"score_gpt":0.2599844484148164,"score_spread":0.2348734288788328,"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."}}