{"id":"W4389520394","doi":"10.18653/v1/2023.emnlp-main.652","title":"Balance Act: Mitigating Hubness in Cross-Modal Retrieval with Query and Gallery Banks","year":2023,"lang":"en","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":9,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Softmax function; Inference; Boosting (machine learning); Information retrieval; Normalization (sociology); Similarity (geometry); Dual (grammatical number); Data mining; Artificial intelligence; Image (mathematics); Deep learning","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":[0.004581484,0.001821657,0.002209879,0.002642765,0.001372023,0.002263838,0.002705594,0.001940961,0.005098785],"category_scores_gemma":[0.01278977,0.0006974397,0.001175432,0.002212683,0.001695387,0.005071864,0.004056308,0.0021169,0.003376715],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001048346,"about_ca_system_score_gemma":0.00177926,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005662705,"about_ca_topic_score_gemma":0.00845239,"domain_scores_codex":[0.997343,0.0005778743,0.0001935792,0.0007727173,0.0008012755,0.0003116294],"domain_scores_gemma":[0.9958254,0.001482968,0.000348447,0.00120072,0.0009631396,0.0001794255],"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":[0.00241947,0.0007356654,0.005977981,0.0005749583,0.0003787304,0.0003792407,0.0009629331,0.05442625,0.08122078,0.009516234,0.01858514,0.8248227],"study_design_scores_gemma":[0.0001446434,0.0005104007,0.005940326,0.00005992451,0.0002964732,0.0005026221,0.0005378568,0.8974007,0.06600963,0.01852829,0.009943168,0.000125936],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06693676,0.00132157,0.9204487,0.0004209595,0.0001592381,0.0003078083,0.0004720539,0.006741966,0.003190899],"genre_scores_gemma":[0.6312526,0.000741894,0.346719,0.001291397,0.0003785478,0.000445609,0.002565288,0.001013376,0.01559224],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005662705,"threshold_uncertainty_score":0.02422947,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01178640720667072,"score_gpt":0.2904348914203238,"score_spread":0.2786484842136531,"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."}}