{"id":"W3200285169","doi":"10.18653/v1/2021.emnlp-main.779","title":"PICARD: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models","year":2021,"lang":"en","type":"preprint","venue":"Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":12,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University; University of Toronto","funders":"","keywords":"Computer science; Decoding methods; Parsing; Language model; Natural language processing; Artificial intelligence; Code (set theory); Programming language; Compiler; Algorithm; Set (abstract data type)","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":["metaepi_narrow","scholarly_communication","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.002074232,0.0009895816,0.001612364,0.000519568,0.0003218084,0.001518421,0.004573331,0.0008734342,0.00002417613],"category_scores_gemma":[0.003512418,0.0007533261,0.0005878755,0.001235644,0.0003693929,0.001215651,0.004061878,0.003076923,5.450121e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005212605,"about_ca_system_score_gemma":0.0009505381,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001812571,"about_ca_topic_score_gemma":0.00002823347,"domain_scores_codex":[0.99413,0.0003214512,0.001414924,0.002082405,0.001084288,0.00096695],"domain_scores_gemma":[0.9949386,0.0009993778,0.001807228,0.0008305313,0.001236328,0.0001878979],"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":[0.0001573801,0.0002309112,0.0002895589,0.001578742,0.0001507389,0.00004285729,0.02253228,0.00007352937,0.2412997,0.004694725,0.00009351339,0.728856],"study_design_scores_gemma":[0.0008674943,0.00007771164,0.00008703404,0.01001715,0.0001438567,0.00002534278,0.004891908,0.5860032,0.317344,0.07935237,0.000008317571,0.001181654],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1111609,0.07104825,0.8051189,0.004927959,0.001763297,0.002864613,0.000145042,0.0008216914,0.002149282],"genre_scores_gemma":[0.4737535,0.00003715678,0.5253844,0.0004296361,0.0001388668,0.0001332786,0.00002539382,0.00004665595,0.00005117034],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.7276744,"threshold_uncertainty_score":0.9995181,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07064381815737095,"score_gpt":0.4144700036335565,"score_spread":0.3438261854761855,"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."}}