{"id":"W4416037156","doi":"10.18653/v1/2025.emnlp-industry.134","title":"Deploying Tiny LVLM Judges for Real-World Evaluation of Chart Models: Lessons Learned and Best Practices","year":2025,"lang":"","type":"article","venue":"","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Compute Canada","keywords":"Best practice; Chart; Natural language; Natural (archaeology); Column (typography)","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":[],"consensus_categories":[],"category_scores_codex":[0.005474941,0.0002440042,0.0003553794,0.0004486991,0.000478382,0.0004578775,0.0005807611,0.0001221123,0.00004372456],"category_scores_gemma":[0.002296406,0.0002377535,0.00007697101,0.0008067989,0.0001112758,0.00161177,0.0003420398,0.0002649886,0.00001009097],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009101921,"about_ca_system_score_gemma":0.0006549957,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001799779,"about_ca_topic_score_gemma":0.001673756,"domain_scores_codex":[0.996776,0.0006370196,0.0006600007,0.0009610439,0.0006372673,0.0003287089],"domain_scores_gemma":[0.9957721,0.001127272,0.001153865,0.0009312918,0.0009254137,0.00009003006],"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":[0.00005573478,0.0002423616,0.002185472,0.0003053628,0.00007498335,1.350019e-7,0.0008313665,0.00368306,0.0007951786,0.3046317,0.001047721,0.686147],"study_design_scores_gemma":[0.0009302818,0.0001436361,0.003659392,0.0002147208,0.0003848387,0.000001047196,0.0005810165,0.969632,0.0006639574,0.01704227,0.006534727,0.0002121052],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02327051,0.002074353,0.7984527,0.06954528,0.0006001274,0.001731148,0.00003612974,0.0001781671,0.1041116],"genre_scores_gemma":[0.9620358,0.001623652,0.02619517,0.0001788439,0.00007469542,0.0001321782,0.00005224524,0.00001706546,0.009690321],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9659489,"threshold_uncertainty_score":0.9695306,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3862536639501394,"score_gpt":0.4808505296975183,"score_spread":0.09459686574737886,"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."}}