{"id":"W4390573228","doi":"10.48550/arxiv.2401.00907","title":"LaFFi: Leveraging Hybrid Natural Language Feedback for Fine-tuning Language Models","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; University of Alberta; Alberta Machine Intelligence Institute; Compute Canada; Mitacs; Canadian Institute for Advanced Research","keywords":"Computer science; Fine-tuning; Realm; Natural language; Natural (archaeology); Language model; Natural language understanding; Task (project management); Artificial intelligence; Natural language processing; Engineering","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":[0.0003734373,0.0004070608,0.0004254302,0.0003625707,0.0002160063,0.0002292696,0.002345976,0.0001804498,0.000008649547],"category_scores_gemma":[0.00006813742,0.0004850168,0.0003263352,0.0003732479,0.00004641006,0.0005801806,0.002742423,0.0007506123,0.00006115899],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002506156,"about_ca_system_score_gemma":0.0001757533,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004180867,"about_ca_topic_score_gemma":0.0001025237,"domain_scores_codex":[0.9973823,0.00008781689,0.0002653908,0.001488209,0.0001482767,0.000627947],"domain_scores_gemma":[0.9977257,0.0002305414,0.0002323144,0.001525438,0.0001319923,0.0001540575],"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":[0.00001915639,0.00002723635,0.00007620947,0.0002018941,0.0001017649,0.0007715726,0.003469504,0.9504365,0.0001959972,0.03994605,0.000594563,0.00415957],"study_design_scores_gemma":[0.000540995,0.00001358069,0.00002928986,0.0001635535,0.00004017348,0.000007776161,0.0005746913,0.9781458,0.0002754568,0.01961206,0.0000727771,0.0005238893],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2832697,0.0002217566,0.7135847,0.0001782382,0.001091328,0.000397206,0.00003114935,0.0007369803,0.0004888586],"genre_scores_gemma":[0.9757831,0.00002542876,0.01586385,0.0001432652,0.000246989,0.000004303176,0.00007129921,0.00005260176,0.007809232],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.6977209,"threshold_uncertainty_score":0.9997602,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1021838348557336,"score_gpt":0.2127921687248835,"score_spread":0.1106083338691499,"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."}}