{"id":"W4413817033","doi":"10.1016/j.engappai.2025.112039","title":"Gradient surgery: A necessity for robust test-time adaptation for detecting casting defects","year":2025,"lang":"en","type":"article","venue":"Engineering Applications of Artificial Intelligence","topic":"Non-Destructive Testing Techniques","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"Natural Sciences and Engineering Research Council of Canada; Concordia University","keywords":"Computer science; Adaptation (eye); Test (biology); Artificial intelligence","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000462012,0.0001811062,0.0002469835,0.0003115611,0.0001353151,0.00003503509,0.0002051578,0.00008295698,0.000001749399],"category_scores_gemma":[0.001739906,0.0002287986,0.0001111364,0.0006501632,0.00003617725,0.00009517585,0.0000278669,0.0001152906,0.000002639782],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001196752,"about_ca_system_score_gemma":0.00003857084,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002236652,"about_ca_topic_score_gemma":0.00001085546,"domain_scores_codex":[0.9988795,0.000006831885,0.0005007805,0.0002553215,0.00008307965,0.0002744828],"domain_scores_gemma":[0.9963974,0.00295388,0.00009634096,0.000274887,0.0002340334,0.00004341431],"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":[0.00002044326,0.0000701472,0.0002093026,0.001181397,0.00004711113,2.363372e-7,0.0002209031,0.6153786,0.203281,0.04393025,0.000100759,0.1355599],"study_design_scores_gemma":[0.00001367583,0.00002276979,0.00008694292,0.0001375718,0.00002850574,0.000001204332,0.00004561801,0.7141778,0.2648346,0.02043138,0.00005210355,0.0001677673],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03787051,0.00007574561,0.9598152,0.00002126575,0.000112456,0.001051296,0.00003770512,0.000927924,0.00008789625],"genre_scores_gemma":[0.5194152,0.000002848624,0.4795523,0.000002735761,0.00004283097,0.0009442993,0.000009676471,0.00002711414,0.000003008812],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.4815447,"threshold_uncertainty_score":0.9330138,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03737551167254137,"score_gpt":0.2685418942224299,"score_spread":0.2311663825498885,"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."}}