{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.005962182,0.001854948,0.001343788,0.00341922,0.0005909362,0.00227133,0.001876449,0.001232353,0.003768239],"category_scores_gemma":[0.04392879,0.000277509,0.0007153112,0.001872483,0.0007464681,0.002333464,0.001679017,0.001081632,0.001860232],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001235177,"about_ca_system_score_gemma":0.0006061497,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002249065,"about_ca_topic_score_gemma":0.002420586,"domain_scores_codex":[0.9927492,0.003198662,0.0006464203,0.001112217,0.00200137,0.0002921065],"domain_scores_gemma":[0.9495935,0.03600958,0.002588197,0.004275494,0.005822095,0.001711182],"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":[0.006099046,0.002856949,0.02168593,0.004686301,0.0007429929,0.0004243249,0.00294323,0.07598969,0.06082315,0.002881095,0.02868055,0.7921869],"study_design_scores_gemma":[0.0006529773,0.009332045,0.05486029,0.0002198226,0.0002780045,0.0006662636,0.001954899,0.8236519,0.08949225,0.005003322,0.01362904,0.0002592261],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.807432,0.003477358,0.1446155,0.0003842297,0.0005369049,0.001921871,0.005577968,0.02210821,0.01394592],"genre_scores_gemma":[0.8823325,0.0004373036,0.09861771,0.0001398666,0.0001256765,0.0008364537,0.01254126,0.001147424,0.003821682],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.005962182,"threshold_uncertainty_score":0.03153139,"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."}}