{"id":"W7138037567","doi":"10.18653/v1/2025.findings-ijcnlp.66","title":"Illusions of Relevance: Arbitrary Content Injection Attacks Deceive Retrievers, Rerankers, and LLM Judges","year":2025,"lang":"","type":"article","venue":"","topic":"Web Application Security Vulnerabilities","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Illusion; Content (measure theory); Perception; Deception","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006305918,0.0006205441,0.001051525,0.00123556,0.002302335,0.003161456,0.001395403,0.003382928,0.004785513],"category_scores_gemma":[0.05428518,0.0005161542,0.0004707917,0.001005844,0.002931455,0.007209392,0.003589915,0.003761522,0.003386795],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001315582,"about_ca_system_score_gemma":0.0009251138,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001382841,"about_ca_topic_score_gemma":0.001997629,"domain_scores_codex":[0.9922773,0.002686436,0.0003525602,0.0005936008,0.003498744,0.0005912764],"domain_scores_gemma":[0.9478292,0.02616902,0.003171586,0.01661173,0.005024842,0.00119351],"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":[0.003994612,0.0007596438,0.0190195,0.0003258148,0.0003283392,0.001685766,0.004438646,0.01909672,0.04773915,0.2198713,0.1450308,0.5377097],"study_design_scores_gemma":[0.0004800764,0.0009440701,0.008866456,0.0001760095,0.0002896351,0.003002557,0.001771076,0.5041704,0.08659091,0.3411421,0.05232282,0.000244051],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5786127,0.005599566,0.243688,0.03734186,0.001400444,0.0004528182,0.0008433309,0.01102959,0.1210317],"genre_scores_gemma":[0.9736218,0.0003018783,0.01432987,0.00144701,0.0002274541,0.00004451371,0.0001313807,0.0002175092,0.009678523],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.006305918,"threshold_uncertainty_score":0.03334928,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02542288503095858,"score_gpt":0.270656908554714,"score_spread":0.2452340235237554,"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."}}