{"id":"W4414317082","doi":"10.20944/preprints202509.1607.v1","title":"Capturing Narrative Semantics from Captions for Relational Scene Abstraction","year":2025,"lang":"en","type":"preprint","venue":"Preprints.org","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Simon Fraser University","funders":"","keywords":"Interpretability; Scene graph; Closed captioning; Semantics (computer science); Graph; Scalability; Natural language; Exploit; Abstraction","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0004730793,0.0003443055,0.0003486711,0.0002371841,0.0004603372,0.00008762547,0.001273523,0.0003906539,0.0001245671],"category_scores_gemma":[0.0004655094,0.0004059732,0.0002391456,0.0002080159,0.00005908599,0.0002481745,0.001911599,0.001263143,0.0003645631],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003042514,"about_ca_system_score_gemma":0.0003942712,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002769252,"about_ca_topic_score_gemma":0.00009510871,"domain_scores_codex":[0.9972671,0.0001255988,0.0005885683,0.00136872,0.0003504667,0.0002995156],"domain_scores_gemma":[0.9967235,0.0005588348,0.0005080597,0.001709756,0.0003852501,0.0001146229],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"observational","study_design_scores_codex":[0.00005267283,0.0006270368,0.3760536,0.0003769387,0.0007630716,0.00000522678,0.02039149,0.4645635,0.008266472,0.1235141,0.0003085554,0.005077461],"study_design_scores_gemma":[0.00032311,0.000005710806,0.6653286,0.0001967807,0.00006230723,0.000001843883,0.00008705681,0.2828863,0.003165409,0.04541697,0.002113448,0.0004124735],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4166035,0.00004379392,0.5722861,0.005455686,0.0009686865,0.001101744,0.0001641939,0.0005130421,0.002863328],"genre_scores_gemma":[0.8822582,0.00001652292,0.115036,0.0001012311,0.0002664321,0.0007566917,0.0003393499,0.00002016438,0.001205394],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4656547,"threshold_uncertainty_score":0.9998392,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1403948854982185,"score_gpt":0.3842923063313395,"score_spread":0.243897420833121,"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."}}