{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:77QPX2JMJADHFNTAXRRC7HCDDU","short_pith_number":"pith:77QPX2JM","schema_version":"1.0","canonical_sha256":"ffe0fbe92c480672b660bc622f9c431d1a408e41b0bb88e9767dcaf1d96254f3","source":{"kind":"arxiv","id":"2501.18636","version":2},"attestation_state":"computed","paper":{"title":"SafeRAG: Benchmarking Security in Retrieval-Augmented Generation of Large Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.CR","authors_text":"Bo Tang, Feiyu Xiong, Hanyu Wang, Jason Zhaoxin Fan, Jiawei Yang, Mengwei Wang, Sensen Zhang, Shichao Song, Simin Niu, Xun Liang, Zhiyu Li","submitted_at":"2025-01-28T17:01:31Z","abstract_excerpt":"The indexing-retrieval-generation paradigm of retrieval-augmented generation (RAG) has been highly successful in solving knowledge-intensive tasks by integrating external knowledge into large language models (LLMs). However, the incorporation of external and unverified knowledge increases the vulnerability of LLMs because attackers can perform attack tasks by manipulating knowledge. In this paper, we introduce a benchmark named SafeRAG designed to evaluate the RAG security. First, we classify attack tasks into silver noise, inter-context conflict, soft ad, and white Denial-of-Service. Next, we"},"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":"2501.18636","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-01-28T17:01:31Z","cross_cats_sorted":["cs.AI","cs.IR"],"title_canon_sha256":"36093f2f1f452c423e3a632e0ce7ed0da6aadff00b6e751ebdd3374ecd03d810","abstract_canon_sha256":"467c56c181140ee64fe3109148613700a8b8a5e2a92f031bf2c26885b125ad68"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:49.052772Z","signature_b64":"Z1wllbriG7IG9sNYMbc1/DzytyPOrhOhw3utwa/lOtjs8tSqezNzubGtjEBdkcneSAyZKSr9esJbkwbV6jFuCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ffe0fbe92c480672b660bc622f9c431d1a408e41b0bb88e9767dcaf1d96254f3","last_reissued_at":"2026-07-05T10:18:49.052178Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:49.052178Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeRAG: Benchmarking Security in Retrieval-Augmented Generation of Large Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.CR","authors_text":"Bo Tang, Feiyu Xiong, Hanyu Wang, Jason Zhaoxin Fan, Jiawei Yang, Mengwei Wang, Sensen Zhang, Shichao Song, Simin Niu, Xun Liang, Zhiyu Li","submitted_at":"2025-01-28T17:01:31Z","abstract_excerpt":"The indexing-retrieval-generation paradigm of retrieval-augmented generation (RAG) has been highly successful in solving knowledge-intensive tasks by integrating external knowledge into large language models (LLMs). However, the incorporation of external and unverified knowledge increases the vulnerability of LLMs because attackers can perform attack tasks by manipulating knowledge. In this paper, we introduce a benchmark named SafeRAG designed to evaluate the RAG security. First, we classify attack tasks into silver noise, inter-context conflict, soft ad, and white Denial-of-Service. Next, we"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.18636","kind":"arxiv","version":2},"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/2501.18636/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":"2501.18636","created_at":"2026-07-05T10:18:49.052244+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.18636v2","created_at":"2026-07-05T10:18:49.052244+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.18636","created_at":"2026-07-05T10:18:49.052244+00:00"},{"alias_kind":"pith_short_12","alias_value":"77QPX2JMJADH","created_at":"2026-07-05T10:18:49.052244+00:00"},{"alias_kind":"pith_short_16","alias_value":"77QPX2JMJADHFNTA","created_at":"2026-07-05T10:18:49.052244+00:00"},{"alias_kind":"pith_short_8","alias_value":"77QPX2JM","created_at":"2026-07-05T10:18:49.052244+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11470","citing_title":"The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes","ref_index":144,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16282","citing_title":"Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05244","citing_title":"Towards Dependable Retrieval-Augmented Generation Using Factual Confidence Prediction","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU","json":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU.json","graph_json":"https://pith.science/api/pith-number/77QPX2JMJADHFNTAXRRC7HCDDU/graph.json","events_json":"https://pith.science/api/pith-number/77QPX2JMJADHFNTAXRRC7HCDDU/events.json","paper":"https://pith.science/paper/77QPX2JM"},"agent_actions":{"view_html":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU","download_json":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU.json","view_paper":"https://pith.science/paper/77QPX2JM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.18636&json=true","fetch_graph":"https://pith.science/api/pith-number/77QPX2JMJADHFNTAXRRC7HCDDU/graph.json","fetch_events":"https://pith.science/api/pith-number/77QPX2JMJADHFNTAXRRC7HCDDU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU/action/storage_attestation","attest_author":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU/action/author_attestation","sign_citation":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU/action/citation_signature","submit_replication":"https://pith.science/pith/77QPX2JMJADHFNTAXRRC7HCDDU/action/replication_record"}},"created_at":"2026-07-05T10:18:49.052244+00:00","updated_at":"2026-07-05T10:18:49.052244+00:00"}