{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:HQJAD6FERC3SLXXGGGQAFMHHGK","short_pith_number":"pith:HQJAD6FE","schema_version":"1.0","canonical_sha256":"3c1201f8a488b725dee631a002b0e73292337632e7363049fac2b2ecddb72010","source":{"kind":"arxiv","id":"2310.19156","version":1},"attestation_state":"computed","paper":{"title":"Poisoning Retrieval Corpora by Injecting Adversarial Passages","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Alexander Wettig, Danqi Chen, Zexuan Zhong, Ziqing Huang","submitted_at":"2023-10-29T21:13:31Z","abstract_excerpt":"Dense retrievers have achieved state-of-the-art performance in various information retrieval tasks, but to what extent can they be safely deployed in real-world applications? In this work, we propose a novel attack for dense retrieval systems in which a malicious user generates a small number of adversarial passages by perturbing discrete tokens to maximize similarity with a provided set of training queries. When these adversarial passages are inserted into a large retrieval corpus, we show that this attack is highly effective in fooling these systems to retrieve them for queries that were not"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2310.19156","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-10-29T21:13:31Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"aed819eb17ca930c0e991fc603a63654aab5f6ab2dc697db2ce505f0b2d840c7","abstract_canon_sha256":"7387413ce78892f03f988325ab3cf245d3fdc8e9e6a82b679ae1f41eaebdd78a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:06:46.804614Z","signature_b64":"eHGZC993/04JydCPZ6NUzw46lppwaQjEOa8+SEMxFRTvyq4eCc9LXn+DOAWtsm+BbC0N9YueOWytCK7OmBpuAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3c1201f8a488b725dee631a002b0e73292337632e7363049fac2b2ecddb72010","last_reissued_at":"2026-07-05T07:06:46.804136Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:06:46.804136Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Poisoning Retrieval Corpora by Injecting Adversarial Passages","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Alexander Wettig, Danqi Chen, Zexuan Zhong, Ziqing Huang","submitted_at":"2023-10-29T21:13:31Z","abstract_excerpt":"Dense retrievers have achieved state-of-the-art performance in various information retrieval tasks, but to what extent can they be safely deployed in real-world applications? In this work, we propose a novel attack for dense retrieval systems in which a malicious user generates a small number of adversarial passages by perturbing discrete tokens to maximize similarity with a provided set of training queries. When these adversarial passages are inserted into a large retrieval corpus, we show that this attack is highly effective in fooling these systems to retrieve them for queries that were not"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.19156","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2310.19156/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2310.19156","created_at":"2026-07-05T07:06:46.804188+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.19156v1","created_at":"2026-07-05T07:06:46.804188+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.19156","created_at":"2026-07-05T07:06:46.804188+00:00"},{"alias_kind":"pith_short_12","alias_value":"HQJAD6FERC3S","created_at":"2026-07-05T07:06:46.804188+00:00"},{"alias_kind":"pith_short_16","alias_value":"HQJAD6FERC3SLXXG","created_at":"2026-07-05T07:06:46.804188+00:00"},{"alias_kind":"pith_short_8","alias_value":"HQJAD6FE","created_at":"2026-07-05T07:06:46.804188+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01276","citing_title":"Embedding Inference Attack","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12290","citing_title":"Selection Integrity for LLM Graph Memory: An Accumulability Criterion for Information-Flow-Blind Retrieval","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27736","citing_title":"ToE: A Hierarchical and Explainable Claim Verification Framework with Dynamic Multi-source Evidence Retrieval and Aggregation","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27157","citing_title":"Detecting Is Not Resolving: The Monitoring Control Gap in Retrieval Augmented LLMs","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2506.04390","citing_title":"Through the Stealth Lens: Attention-Aware Defenses Against Poisoning in RAG","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2508.03793","citing_title":"AttnTrace: Contextual Attribution of Prompt Injection and Knowledge Corruption","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2510.23883","citing_title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01591","citing_title":"Led to Mislead: Adversarial Content Injection for Attacks on Neural Ranking Models","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01782","citing_title":"Needle-in-RAG: Prompt-Conditioned Character-Level Traceback of Poisoned Spans in Retrieved Evidence","ref_index":66,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK","json":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK.json","graph_json":"https://pith.science/api/pith-number/HQJAD6FERC3SLXXGGGQAFMHHGK/graph.json","events_json":"https://pith.science/api/pith-number/HQJAD6FERC3SLXXGGGQAFMHHGK/events.json","paper":"https://pith.science/paper/HQJAD6FE"},"agent_actions":{"view_html":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK","download_json":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK.json","view_paper":"https://pith.science/paper/HQJAD6FE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.19156&json=true","fetch_graph":"https://pith.science/api/pith-number/HQJAD6FERC3SLXXGGGQAFMHHGK/graph.json","fetch_events":"https://pith.science/api/pith-number/HQJAD6FERC3SLXXGGGQAFMHHGK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK/action/storage_attestation","attest_author":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK/action/author_attestation","sign_citation":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK/action/citation_signature","submit_replication":"https://pith.science/pith/HQJAD6FERC3SLXXGGGQAFMHHGK/action/replication_record"}},"created_at":"2026-07-05T07:06:46.804188+00:00","updated_at":"2026-07-05T07:06:46.804188+00:00"}