{"id":"W4313549837","doi":"10.1145/3551349.3556900","title":"AST-Probe: Recovering abstract syntax trees from hidden representations of pre-trained language models","year":2022,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"","keywords":"Computer science; ENCODE; Natural language processing; Syntax; Artificial intelligence; Language model; Subspace topology; Representation (politics); Abstract syntax tree; Natural language","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":[0.000180437,0.00008849988,0.0001396001,0.00008168425,0.0001058497,0.0000519464,0.0007701714,0.00002259994,0.0004585065],"category_scores_gemma":[0.00002780352,0.00009039398,0.00006921632,0.0001839425,0.00001165483,0.0004900214,0.0005200062,0.0001294614,0.000002823011],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000463784,"about_ca_system_score_gemma":0.00005510845,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003865991,"about_ca_topic_score_gemma":0.0001862976,"domain_scores_codex":[0.9988169,0.00004131913,0.0002870261,0.00036286,0.0003301561,0.000161719],"domain_scores_gemma":[0.9990093,0.0001191428,0.0001063267,0.0006936687,0.00002530038,0.00004631007],"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":[0.00004681837,0.0003322405,0.001073693,0.00004177858,0.0001634832,0.00008783405,0.04113516,0.5754814,0.09706851,0.1041784,0.001144032,0.1792467],"study_design_scores_gemma":[0.0002893405,0.00003395235,0.005088088,0.0000125062,0.000007901769,0.000005949625,0.001408784,0.9617403,0.005187641,0.02599553,0.00005858077,0.0001714878],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.407644,0.00007288979,0.5838613,0.0003406105,0.0001608472,0.000135066,0.00002823804,0.000148579,0.007608484],"genre_scores_gemma":[0.8392536,0.000002215105,0.159532,0.00006696826,0.0000364624,0.00003223554,0.00001373052,0.000005762992,0.001057055],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4316096,"threshold_uncertainty_score":0.5844247,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02802503178024772,"score_gpt":0.2726645080868488,"score_spread":0.2446394763066011,"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."}}