{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:V5PHTMD3BJIOBAU4TD3IUX3RUN","short_pith_number":"pith:V5PHTMD3","schema_version":"1.0","canonical_sha256":"af5e79b07b0a50e0829c98f68a5f71a3729dd9fb30a1514d762c6bcb0698aec8","source":{"kind":"arxiv","id":"2505.06579","version":1},"attestation_state":"computed","paper":{"title":"POISONCRAFT: Practical Poisoning of Retrieval-Augmented Generation for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Chengshang Hou, Gang Xiong, Haozheng Luo, Jiahao Yu, Junzheng Shi, Xinjie Lin, Yangguang Shao","submitted_at":"2025-05-10T09:36:28Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable success in various domains, primarily due to their strong capabilities in reasoning and generating human-like text. Despite their impressive performance, LLMs are susceptible to hallucinations, which can lead to incorrect or misleading outputs. This is primarily due to the lack of up-to-date knowledge or domain-specific information. Retrieval-augmented generation (RAG) is a promising approach to mitigate hallucinations by leveraging external knowledge sources. However, the security of RAG systems has not been thoroughly studied. In this pap"},"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":"2505.06579","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-05-10T09:36:28Z","cross_cats_sorted":[],"title_canon_sha256":"9b389c0b8ad7cc0bb9e140c67301a609ac73cceead24586fba6b0f4d17d23b6d","abstract_canon_sha256":"ca418fd4b412a5044ec99c86e21de27cbec7ef3b399d559cc3ebbabf07729fa5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:01:12.550350Z","signature_b64":"ib95adKqkXncvgz7UN6lfhJLalb7oz4IhI1ztzCz9gVLusMBfD6t3wUARdXxHImpiLwmY4v7QzGwSDTOpO1VAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"af5e79b07b0a50e0829c98f68a5f71a3729dd9fb30a1514d762c6bcb0698aec8","last_reissued_at":"2026-07-05T11:01:12.549822Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:01:12.549822Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"POISONCRAFT: Practical Poisoning of Retrieval-Augmented Generation for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Chengshang Hou, Gang Xiong, Haozheng Luo, Jiahao Yu, Junzheng Shi, Xinjie Lin, Yangguang Shao","submitted_at":"2025-05-10T09:36:28Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable success in various domains, primarily due to their strong capabilities in reasoning and generating human-like text. Despite their impressive performance, LLMs are susceptible to hallucinations, which can lead to incorrect or misleading outputs. This is primarily due to the lack of up-to-date knowledge or domain-specific information. Retrieval-augmented generation (RAG) is a promising approach to mitigate hallucinations by leveraging external knowledge sources. However, the security of RAG systems has not been thoroughly studied. In this pap"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.06579","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/2505.06579/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":"2505.06579","created_at":"2026-07-05T11:01:12.549887+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.06579v1","created_at":"2026-07-05T11:01:12.549887+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.06579","created_at":"2026-07-05T11:01:12.549887+00:00"},{"alias_kind":"pith_short_12","alias_value":"V5PHTMD3BJIO","created_at":"2026-07-05T11:01:12.549887+00:00"},{"alias_kind":"pith_short_16","alias_value":"V5PHTMD3BJIOBAU4","created_at":"2026-07-05T11:01:12.549887+00:00"},{"alias_kind":"pith_short_8","alias_value":"V5PHTMD3","created_at":"2026-07-05T11:01:12.549887+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07783","citing_title":"Evaluating RAG Reliability under Clean, Misleading, and Mixed Retrieval","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28074","citing_title":"SilentRetrieval: Hijacking Retrieval-Augmented Generation via Semantically-Preserving Adversarial Data Poisoning","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2510.10073","citing_title":"SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN","json":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN.json","graph_json":"https://pith.science/api/pith-number/V5PHTMD3BJIOBAU4TD3IUX3RUN/graph.json","events_json":"https://pith.science/api/pith-number/V5PHTMD3BJIOBAU4TD3IUX3RUN/events.json","paper":"https://pith.science/paper/V5PHTMD3"},"agent_actions":{"view_html":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN","download_json":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN.json","view_paper":"https://pith.science/paper/V5PHTMD3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.06579&json=true","fetch_graph":"https://pith.science/api/pith-number/V5PHTMD3BJIOBAU4TD3IUX3RUN/graph.json","fetch_events":"https://pith.science/api/pith-number/V5PHTMD3BJIOBAU4TD3IUX3RUN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN/action/storage_attestation","attest_author":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN/action/author_attestation","sign_citation":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN/action/citation_signature","submit_replication":"https://pith.science/pith/V5PHTMD3BJIOBAU4TD3IUX3RUN/action/replication_record"}},"created_at":"2026-07-05T11:01:12.549887+00:00","updated_at":"2026-07-05T11:01:12.549887+00:00"}