{"id":"W4386827564","doi":"10.1016/j.cose.2023.103482","title":"Transferable adversarial distribution learning: Query-efficient adversarial attack against large language models","year":2023,"lang":"en","type":"article","venue":"Computers & Security","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":16,"is_retracted":false,"has_abstract":false,"ca_institutions":"Brock University","funders":"National Natural Science Foundation of China","keywords":"Computer science; Overfitting; Machine learning; Language model; Artificial intelligence; Leverage (statistics); Regularization (linguistics); Adversarial system; Black box; Artificial neural network","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":[0.001503116,0.0005092872,0.0005642578,0.0002563464,0.0008761312,0.0003565151,0.001734973,0.0002980515,0.00002716241],"category_scores_gemma":[0.0002046641,0.0005513647,0.0003577154,0.00159628,0.0001149888,0.0009027218,0.001120032,0.001252666,0.0002477653],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003260428,"about_ca_system_score_gemma":0.0001972375,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000106608,"about_ca_topic_score_gemma":0.00002000684,"domain_scores_codex":[0.9953467,0.0006479961,0.0005986062,0.001209631,0.0009306322,0.001266416],"domain_scores_gemma":[0.998,0.0003931451,0.000209495,0.000909657,0.0001592366,0.0003285096],"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":[0.0000629591,0.0001419686,0.0001037623,0.00004421523,0.00005752589,0.0001750757,0.008713703,0.9593881,0.00009140879,0.02207432,0.004887242,0.00425973],"study_design_scores_gemma":[0.002312205,0.0001011113,0.0002093295,0.00005795177,0.00002395621,0.000009341307,0.000337566,0.9801105,0.0001055046,0.0008161739,0.01532395,0.0005923685],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1620036,0.00007715099,0.8301702,0.000843987,0.003560271,0.000447832,0.00005924939,0.001993748,0.0008439381],"genre_scores_gemma":[0.9946965,0.00003331073,0.003464089,0.0002276258,0.0007826703,0.00002350875,0.0005953584,0.00004844426,0.000128518],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8326929,"threshold_uncertainty_score":0.9996938,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01381903665489724,"score_gpt":0.2631415999348444,"score_spread":0.2493225632799472,"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."}}