{"id":"W4385571036","doi":"10.18653/v1/2023.findings-acl.120","title":"Pragmatic Inference with a CLIP Listener for Contrastive Captioning","year":2023,"lang":"en","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Samsung","keywords":"Closed captioning; Discriminative model; Computer science; Leverage (statistics); Artificial intelligence; Inference; Fluency; Speech recognition; Natural language processing; Prosody; Hyperparameter; Image (mathematics); Linguistics","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":[0.003496771,0.001868607,0.0009611255,0.001026727,0.0009528279,0.002603632,0.002679591,0.002166436,0.01061782],"category_scores_gemma":[0.01794056,0.0008504369,0.001343099,0.0005388044,0.001783428,0.003811661,0.00298954,0.003723647,0.003245953],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001403581,"about_ca_system_score_gemma":0.001210557,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002289178,"about_ca_topic_score_gemma":0.003214849,"domain_scores_codex":[0.9967554,0.001377325,0.0001170844,0.0009082443,0.0006638397,0.0001781459],"domain_scores_gemma":[0.995191,0.002789306,0.0002871373,0.0008517722,0.0006608975,0.0002199701],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.001017317,0.0003762697,0.002246405,0.0008979692,0.0003310924,0.0007655935,0.002302025,0.1276238,0.07912114,0.1106536,0.04038267,0.6342822],"study_design_scores_gemma":[0.00005771354,0.0001118334,0.0004618592,0.0000416244,0.00006237555,0.0002157448,0.0001434423,0.9143553,0.02232701,0.05049342,0.0116612,0.00006859865],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.006231255,0.0001981915,0.9839094,0.00031241,0.0001186034,0.0001414055,0.0002035185,0.004427598,0.004457548],"genre_scores_gemma":[0.3269249,0.0002008984,0.6624739,0.0009640743,0.0003374226,0.0004084472,0.001263505,0.00137245,0.006054441],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01061782,"threshold_uncertainty_score":0.03552014,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01852115849155336,"score_gpt":0.3110642857477515,"score_spread":0.2925431272561981,"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."}}