{"id":"W4409361367","doi":"10.1609/aaai.v39i26.34929","title":"Scaling Trends for Data Poisoning in LLMs","year":2025,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Scientific Computing and Data Management","field":"Decision Sciences","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Institut de Valorisation des Données","keywords":"Scaling; Environmental health; Environmental science; Medicine; Mathematics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.006535014,0.0001626721,0.0003011715,0.0006854436,0.0002499632,0.0007327183,0.005311564,0.00005907552,0.0001037706],"category_scores_gemma":[0.01054612,0.0001110932,0.00009637373,0.002703232,0.0002289282,0.0004014173,0.001649841,0.0001883681,0.00004029697],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000383728,"about_ca_system_score_gemma":0.0000839321,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008925242,"about_ca_topic_score_gemma":0.00008626133,"domain_scores_codex":[0.9967098,0.0000250066,0.001043461,0.001037082,0.0008381135,0.0003465074],"domain_scores_gemma":[0.9967818,0.00100034,0.0004229636,0.001113262,0.000631084,0.00005049391],"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":[0.00007169607,0.00008496558,0.0009927327,0.00001411373,0.000007306619,1.132263e-7,0.0003461971,0.0001987948,0.001359646,0.3742768,0.006424688,0.6162229],"study_design_scores_gemma":[0.00006433637,0.00005112682,0.001224064,0.0004221437,0.00001858356,4.595275e-7,0.003613283,0.57329,0.03065396,0.3837291,0.006739574,0.0001933766],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5967687,0.0001588958,0.1705943,0.04320587,0.007807042,0.001890193,0.0002160974,0.0002166724,0.1791422],"genre_scores_gemma":[0.9938101,0.000005168417,0.002656308,0.0002061038,0.0000427688,0.00001484095,0.000004324071,0.000005775469,0.003254589],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6160295,"threshold_uncertainty_score":0.9977885,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4474056700080657,"score_gpt":0.476595572378622,"score_spread":0.02918990237055635,"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."}}