{"id":"W4385573873","doi":"10.18653/v1/2022.emnlp-main.586","title":"Contrastive Learning with Expectation-Maximization for Weakly Supervised Phrase Grounding","year":2022,"lang":"en","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"Fundamental Research Funds for the Central Universities; State Key Laboratory of Software Development Environment; Leverhulme Trust","keywords":"Phrase; Computer science; Margin (machine learning); Artificial intelligence; Annotation; Object (grammar); Natural language processing; Perspective (graphical); Expectation–maximization algorithm; Maximization; Pattern recognition (psychology); Speech recognition; Machine learning; Maximum likelihood; Mathematics","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":[0.0002431695,0.0001075702,0.0001095054,0.0001211958,0.000952167,0.0001241577,0.0004048538,0.00001533759,0.0001152375],"category_scores_gemma":[0.0001098029,0.0001024011,0.0000351153,0.0004934811,0.00001853392,0.0003213308,0.0001246047,0.0001915164,0.000008987635],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001034906,"about_ca_system_score_gemma":0.00006953641,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001758617,"about_ca_topic_score_gemma":0.0000100416,"domain_scores_codex":[0.9989042,0.0001167306,0.0001522303,0.0003745497,0.0002481439,0.0002041712],"domain_scores_gemma":[0.9991745,0.0003380985,0.0001038799,0.0002193169,0.0001053121,0.00005892794],"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":[0.0001181677,0.0002487887,0.02518434,0.00002298067,0.00006835693,0.000005030418,0.01000786,0.5806776,0.006012064,0.3349842,0.0002757218,0.04239487],"study_design_scores_gemma":[0.0007982721,0.0002416185,0.01098812,0.000003423174,0.000008040012,0.00000992474,0.001257146,0.9847247,0.0001860742,0.0006027695,0.001001567,0.0001783261],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.09660258,0.000008548746,0.9000764,0.001200617,0.00005563721,0.0004881945,0.000002579603,0.0003728646,0.001192504],"genre_scores_gemma":[0.8449617,4.091522e-7,0.1536762,0.0001544248,0.00002745344,0.0007451896,0.00004362196,0.00001509962,0.0003758445],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7483591,"threshold_uncertainty_score":0.7323394,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.008977920109588727,"score_gpt":0.2370852657185742,"score_spread":0.2281073456089855,"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."}}