{"id":"W7106288924","doi":"10.2139/ssrn.5763184","title":"Efficient Inference Using Large Language Models with Limited Human Data: Fine-Tuning then Rectification","year":2025,"lang":"","type":"preprint","venue":"SSRN Electronic Journal","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Quest University Canada","funders":"","keywords":"Rectification; Inference; Language model; Data modeling; Natural language","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":[0.008381166,0.00144524,0.002968007,0.00153679,0.001238873,0.002973716,0.004103714,0.003512056,0.007290237],"category_scores_gemma":[0.03860902,0.00199077,0.002112746,0.001949377,0.001511908,0.005569458,0.003841504,0.006401495,0.004239972],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001022821,"about_ca_system_score_gemma":0.003018314,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01192902,"about_ca_topic_score_gemma":0.01750247,"domain_scores_codex":[0.9962794,0.001570667,0.0002529635,0.001273591,0.0003204233,0.0003029113],"domain_scores_gemma":[0.9820539,0.01359229,0.0004691673,0.002887752,0.0007178735,0.0002789658],"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":[0.001506064,0.0003614463,0.004834687,0.000586348,0.0005107538,0.00035621,0.0008109052,0.1794067,0.0163283,0.01901069,0.01680171,0.7594862],"study_design_scores_gemma":[0.00009042906,0.00004708242,0.000728043,0.00002747372,0.00007923234,0.00009221561,0.00008800995,0.9509705,0.003017686,0.04270018,0.002131118,0.00002809693],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01524239,0.0008519496,0.9774585,0.001126471,0.0001214473,0.00007585184,0.0003731586,0.00399532,0.0007549507],"genre_scores_gemma":[0.4378211,0.0007983557,0.5489336,0.001149381,0.0006599762,0.0003911089,0.002925772,0.001400798,0.005919978],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01192902,"threshold_uncertainty_score":0.0443244,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06451865640085883,"score_gpt":0.3251743851199639,"score_spread":0.2606557287191051,"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."}}