{"id":"W7106846313","doi":"10.48448/pd7s-0n44","title":"Reference-Guided Verdict: LLMs-as-Judges in Automatic Evaluation of Free-Form QA","year":2025,"lang":"","type":"other","venue":"Open MIND","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Reliability (semiconductor); Process (computing); Semantics (computer science); Generative grammar; Key (lock); Evaluation methods","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":[0.02815506,0.001442163,0.001378529,0.002119569,0.001514874,0.003310158,0.003760878,0.003606691,0.006007165],"category_scores_gemma":[0.1357518,0.0006814311,0.0006842613,0.000826151,0.002535723,0.005043038,0.007174772,0.003120651,0.003672931],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001696015,"about_ca_system_score_gemma":0.002165043,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003612763,"about_ca_topic_score_gemma":0.00589742,"domain_scores_codex":[0.9370012,0.04823101,0.00184612,0.004552464,0.007376767,0.0009924325],"domain_scores_gemma":[0.9279441,0.04000209,0.004337708,0.01259311,0.0126219,0.002501111],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.003640517,0.0008788711,0.01503731,0.0009021772,0.0003110591,0.001121459,0.01129234,0.08594789,0.03637436,0.0468362,0.03847885,0.759179],"study_design_scores_gemma":[0.0001959197,0.000683287,0.00299949,0.000184767,0.00006087197,0.0004023017,0.0009012665,0.9114693,0.02871699,0.04327473,0.01090129,0.0002099051],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.07982864,0.0006389873,0.8953444,0.001098369,0.000248217,0.0006321219,0.0004157641,0.01271332,0.009080248],"genre_scores_gemma":[0.729143,0.00007060304,0.2624733,0.0004808386,0.0001206673,0.0004176967,0.0007108439,0.001441739,0.005141302],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.02815506,"threshold_uncertainty_score":0.1489,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.141849659331139,"score_gpt":0.4095148714994745,"score_spread":0.2676652121683354,"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."}}