{"id":"W4396790742","doi":"10.1109/satml59370.2024.00023","title":"Indiscriminate Data Poisoning Attacks on Pre-trained Feature Extractors","year":2024,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Waterloo","funders":"","keywords":"Computer science; Feature (linguistics); Machine learning; Artificial intelligence; Extractor; Set (abstract data type); Labeled data; Downstream (manufacturing); Training set; Feature vector; Supervised learning; Process (computing); Artificial neural network; Engineering","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":[0.000554301,0.0002313505,0.0001684709,0.0001944974,0.0001725699,0.0006924259,0.002599622,0.0001262635,0.0001032391],"category_scores_gemma":[0.0002612609,0.0001802846,0.00005531267,0.0005345361,0.00003865728,0.001573009,0.001278754,0.0008040213,0.000146663],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006104311,"about_ca_system_score_gemma":0.00009536737,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003961872,"about_ca_topic_score_gemma":0.00001371507,"domain_scores_codex":[0.9979956,0.00009863193,0.0001804445,0.0009212616,0.0004208505,0.000383256],"domain_scores_gemma":[0.9979904,0.0003644638,0.00005011757,0.001467548,0.00002377716,0.0001036711],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0000379915,0.0001193015,0.001144294,0.0002073642,0.0001761085,0.0008513928,0.007586165,0.0186598,0.001136627,0.4335839,0.07165993,0.4648372],"study_design_scores_gemma":[0.0001907211,0.00008915251,0.004528224,0.0001957225,0.00002057732,0.00004296582,0.0001024332,0.9513974,0.0002485437,0.001397935,0.04138938,0.0003969662],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.007008457,0.0002168963,0.9609374,0.006769773,0.001963922,0.0001796599,0.000009515785,0.001373987,0.02154042],"genre_scores_gemma":[0.9040455,0.000007034271,0.08741525,0.0004361283,0.0003364352,0.000005010281,0.0000514876,0.00003272762,0.007670468],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9327376,"threshold_uncertainty_score":0.7351792,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03982274016507497,"score_gpt":0.3439346978512736,"score_spread":0.3041119576861986,"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."}}