{"id":"W4416260937","doi":"10.1007/978-981-95-4109-6_33","title":"Semantic Consistency Guided Backdoor Trigger Inversion and InitDistillNet for Backdoor Detection","year":2025,"lang":"en","type":"book-chapter","venue":"Communications in computer and information science","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Novelis (Canada)","funders":"","keywords":"Backdoor; Inversion (geology); Outlier; Anomaly detection; Consistency (knowledge bases)","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001359845,0.0002692193,0.0003408232,0.001176502,0.000930713,0.0005983271,0.002096683,0.0001897035,0.000005473474],"category_scores_gemma":[0.0002980875,0.0002766397,0.00005941447,0.0004403816,0.0008503413,0.004645428,0.002755253,0.0004594261,0.0000161533],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001708991,"about_ca_system_score_gemma":0.0003607253,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002130122,"about_ca_topic_score_gemma":0.00001641368,"domain_scores_codex":[0.9981569,0.00005640664,0.000782818,0.0004130991,0.0003377241,0.0002530708],"domain_scores_gemma":[0.9966887,0.000675439,0.0004375374,0.001565292,0.000531796,0.0001012529],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000009002159,0.00001252051,0.00004602216,0.0001599552,0.00001235469,3.140524e-7,0.001108969,0.0005099238,0.000008568012,0.5435863,0.0008004673,0.4537457],"study_design_scores_gemma":[0.000612379,0.00006162205,0.0003478712,0.0002650433,0.00001659502,0.00001839238,0.00002319045,0.8723811,0.0000171305,0.01003465,0.115908,0.0003139873],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.00006253844,0.0002843459,0.9353829,0.0013356,0.0005685054,0.0007460214,0.00001255245,0.0001075724,0.06149998],"genre_scores_gemma":[0.1990201,0.004313712,0.7842386,0.003693342,0.0001452767,0.0001650234,0.0001770454,0.00003497064,0.008211942],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.8718712,"threshold_uncertainty_score":0.9999686,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03493198983547104,"score_gpt":0.3011699993093499,"score_spread":0.2662380094738789,"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."}}