{"id":"W4404304671","doi":"10.1145/3673791.3698424","title":"Offline Evaluation of Set-Based Text-to-Image Generation","year":2024,"lang":"en","type":"preprint","venue":"","topic":"Video Analysis and Summarization","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Set (abstract data type); Benchmark (surveying); Ranking (information retrieval); Process (computing); Similarity (geometry); Image (mathematics); Artificial intelligence; Machine learning; Information retrieval; Data mining; Data science","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":[],"consensus_categories":[],"category_scores_codex":[0.002004641,0.0001726933,0.0002621009,0.000444966,0.00003683812,0.0003085319,0.0005288542,0.000145879,0.0001576212],"category_scores_gemma":[0.0001504647,0.0001499676,0.0001655525,0.0005840183,0.00001224376,0.0001051948,0.0007187842,0.0001763104,0.0001211736],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001176869,"about_ca_system_score_gemma":0.0005571319,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001207918,"about_ca_topic_score_gemma":0.0001791274,"domain_scores_codex":[0.9975392,0.0002172724,0.000501137,0.0006213014,0.001003021,0.0001180298],"domain_scores_gemma":[0.9979209,0.00002752425,0.0001684843,0.0008564195,0.0009666334,0.00006002163],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000004107178,0.0001232236,0.00008752772,0.0002425383,0.0001569802,0.000002308278,0.0005694642,0.756614,0.0343946,0.02979791,0.0198901,0.1581172],"study_design_scores_gemma":[0.0001066204,0.00002316205,0.0001586779,0.00005247085,0.0001430052,1.591067e-7,0.000004084953,0.9763285,0.0165757,0.006246461,0.0002077682,0.0001533747],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01206937,0.0001405759,0.9795427,0.002736721,0.0006137023,0.0004184517,0.00001345047,0.0001111717,0.004353893],"genre_scores_gemma":[0.8903389,0.000005578152,0.1081592,0.0003560493,0.0002021417,0.00006377712,0.0003377408,0.00001313963,0.0005234168],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8782696,"threshold_uncertainty_score":0.6115502,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07140135182829305,"score_gpt":0.3354684726010519,"score_spread":0.2640671207727588,"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."}}