{"id":"W4285288434","doi":"10.18653/v1/2022.acl-long.233","title":"LAGr: Label Aligned Graphs for Better Systematic Generalization in Semantic Parsing","year":2022,"lang":"en","type":"article","venue":"Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute; McGill University; Canadian Institute for Advanced Research","funders":"Samsung; Canadian Institute for Advanced Research; Microsoft Research","keywords":"Computer science; Parsing; Generalization; Artificial intelligence; Natural language processing; Inference; Graph; Sequence (biology); Theoretical computer science; Mathematics","routes":{"ca_aff":true,"ca_fund":true,"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":[0.003422579,0.002129575,0.001080717,0.001909436,0.001292361,0.001627258,0.002674987,0.002622894,0.007071542],"category_scores_gemma":[0.0128982,0.001204298,0.002252055,0.001699328,0.002110155,0.007447131,0.003773266,0.004998517,0.004437056],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001137721,"about_ca_system_score_gemma":0.002250053,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003758499,"about_ca_topic_score_gemma":0.0105855,"domain_scores_codex":[0.9970205,0.001280471,0.0001127564,0.001079083,0.0003895805,0.0001174748],"domain_scores_gemma":[0.9925119,0.004156501,0.000393822,0.0021678,0.0006069339,0.0001630154],"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":[0.0005454,0.0004426655,0.002545456,0.000642047,0.0002599313,0.0009060171,0.001786633,0.1588454,0.03121144,0.1145523,0.05214482,0.6361179],"study_design_scores_gemma":[0.00007794424,0.00007896793,0.0004608237,0.00006396568,0.00005908957,0.0002158883,0.000186248,0.6642701,0.01242969,0.3064728,0.01562104,0.00006348714],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.004696704,0.0001124967,0.9792909,0.0003049261,0.00005201177,0.00007756748,0.0005560963,0.0139561,0.0009531537],"genre_scores_gemma":[0.1008379,0.0001983471,0.8866203,0.0007161041,0.00008204072,0.0002775079,0.005001573,0.003638598,0.002627576],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007071542,"threshold_uncertainty_score":0.02365673,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.008235714887199415,"score_gpt":0.2424025283105196,"score_spread":0.2341668134233202,"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."}}