{"id":"W2952903800","doi":"10.1109/icst.2019.00019","title":"BugsJS: a Benchmark of JavaScript Bugs","year":2019,"lang":"en","type":"article","venue":"","topic":"Software Testing and Debugging Techniques","field":"Computer Science","cited_by":94,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"JavaScript; Unobtrusive JavaScript; Computer science; Benchmark (surveying); Programming language; Test case; Software bug; Unit testing; Rich Internet application; Operating system; Database; Software; Machine 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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002333794,0.00007520448,0.0001282834,0.00008144735,0.00001817591,0.00003204594,0.0006472999,0.00003796847,0.00008091523],"category_scores_gemma":[0.0000950806,0.00006206841,0.00004992602,0.0002561047,0.00001950653,0.000178754,0.0001981764,0.00005767579,0.00009907159],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001055945,"about_ca_system_score_gemma":0.00003579499,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001530691,"about_ca_topic_score_gemma":9.568513e-7,"domain_scores_codex":[0.9992791,0.00001986787,0.0001544409,0.0002158117,0.0001820731,0.0001487426],"domain_scores_gemma":[0.9990231,0.000150157,0.00005727322,0.0006584922,0.00007230979,0.00003865157],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.000009027037,0.0002456314,0.1716663,0.0001394588,0.00004434886,0.00001276474,0.00124866,0.00002521184,0.00544961,0.3021294,0.3956679,0.1233618],"study_design_scores_gemma":[0.001326701,0.001481453,0.05134641,0.00048681,0.00002030144,0.0001378466,0.00003691667,0.1305384,0.0793455,0.694615,0.03922489,0.001439716],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07177263,0.0001738506,0.8034472,0.00052005,0.0006268471,0.0002478593,9.780554e-7,0.01359485,0.1096157],"genre_scores_gemma":[0.675952,0.000001874927,0.3225239,0.0002777603,0.00001057279,0.000003368389,4.246464e-7,0.000003838832,0.001226235],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6041794,"threshold_uncertainty_score":0.2531076,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.004992930306730839,"score_gpt":0.2021300423044523,"score_spread":0.1971371119977214,"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."}}