{"id":"W4416037383","doi":"10.18653/v1/2025.emnlp-demos.68","title":"From Behavioral Performance to Internal Competence: Interpreting Vision-Language Models with VLM-Lens","year":2025,"lang":"","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Natural (archaeology); Empirical research; Natural language; Action (physics)","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":[0.000391113,0.0005447813,0.0005098749,0.0004146825,0.0005140592,0.0009113931,0.002887352,0.0001520244,0.0003609273],"category_scores_gemma":[0.00003806476,0.000478161,0.000125867,0.001259253,0.0001847401,0.001204308,0.002149303,0.001024192,0.0003751219],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002339693,"about_ca_system_score_gemma":0.0002816942,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01395402,"about_ca_topic_score_gemma":0.0002949886,"domain_scores_codex":[0.9961337,0.000168237,0.0007791689,0.001531223,0.0006358751,0.000751752],"domain_scores_gemma":[0.99717,0.0002720962,0.0002363433,0.00173395,0.0002909531,0.0002966592],"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":[0.000389051,0.001153029,0.0529927,0.0001054923,0.0001819896,0.00004999443,0.05259276,0.1745439,0.007749609,0.02044984,0.001050189,0.6887414],"study_design_scores_gemma":[0.0005827738,0.000542687,0.02156436,0.001542429,0.00005003731,0.000009386652,0.0007879401,0.9725003,0.001420417,0.0002679727,0.0001945651,0.0005371887],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4811887,0.00005820226,0.5051317,0.002035012,0.0003048677,0.0004080034,0.00001534168,0.0002025422,0.01065565],"genre_scores_gemma":[0.8003623,0.000008722329,0.196296,0.001455738,0.00008907589,0.00008621164,0.0000105229,0.00002776074,0.001663558],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7979563,"threshold_uncertainty_score":0.999767,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01081601759030118,"score_gpt":0.3170005470036598,"score_spread":0.3061845294133586,"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."}}