{"id":"W4398766359","doi":"10.1145/3639476.3639774","title":"Naturalness of Attention: Revisiting Attention in Code Language Models","year":2024,"lang":"en","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Computer science; Interpretability; Naturalness; Python (programming language); Artificial intelligence; Java; Source code; Programming language; Empirical research; Natural language processing","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003348529,0.0005808054,0.0005058026,0.001022007,0.00048541,0.001910475,0.00152,0.0007903158,0.002165138],"category_scores_gemma":[0.03196429,0.000449481,0.0008454975,0.0005742093,0.00299407,0.00606738,0.002104491,0.002627387,0.0001879778],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001757639,"about_ca_system_score_gemma":0.001047635,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008898186,"about_ca_topic_score_gemma":0.006714337,"domain_scores_codex":[0.9984554,0.0008717219,0.00005092864,0.0003128392,0.0002038504,0.0001052006],"domain_scores_gemma":[0.9851795,0.01181843,0.0007214937,0.001303724,0.0007092892,0.0002676739],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0001946963,0.00009376716,0.01086815,0.0002139698,0.0001138223,0.0002248652,0.002764226,0.2613508,0.006706854,0.6219961,0.001401128,0.09407157],"study_design_scores_gemma":[0.000007664002,0.00003014584,0.001420276,0.00002671996,0.00001495733,0.00004165922,0.0001054823,0.6283497,0.0008714864,0.3682444,0.0008753192,0.0000122697],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1497848,0.0003541092,0.8413575,0.002357527,0.00004916082,0.00004608315,0.0001299446,0.0003560427,0.005564845],"genre_scores_gemma":[0.9501144,0.0002365029,0.04789231,0.0002674659,0.00006484408,0.00008075905,0.00009962259,0.0001504077,0.00109357],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.008898186,"threshold_uncertainty_score":0.01770896,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02059766564120435,"score_gpt":0.2981587685741326,"score_spread":0.2775611029329282,"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."}}