{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:DGCQMFEIV423BMJQEQ2T4DAWPJ","short_pith_number":"pith:DGCQMFEI","schema_version":"1.0","canonical_sha256":"1985061488af35b0b13024353e0c167a43775205c5aa2876e51aa725408c3448","source":{"kind":"arxiv","id":"2402.07179","version":3},"attestation_state":"computed","paper":{"title":"Prompt Perturbation in Retrieval-Augmented Generation based Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Chen Wang, Helen (Hye-Young) Paik, Liming Zhu, Yanfeng Shu, Zhibo Hu","submitted_at":"2024-02-11T12:25:41Z","abstract_excerpt":"The robustness of large language models (LLMs) becomes increasingly important as their use rapidly grows in a wide range of domains. Retrieval-Augmented Generation (RAG) is considered as a means to improve the trustworthiness of text generation from LLMs. However, how the outputs from RAG-based LLMs are affected by slightly different inputs is not well studied. In this work, we find that the insertion of even a short prefix to the prompt leads to the generation of outputs far away from factually correct answers. We systematically evaluate the effect of such prefixes on RAG by introducing a nov"},"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":"2402.07179","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-11T12:25:41Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"105f23a8434f5773ecaebed7f4a87f3cdafa9401d624e7694f230fcd1e79e4cc","abstract_canon_sha256":"8b52c3956a37a0aa3ee7fa1a75a12033d995dd237ed4e704f5417c34d9e8103b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:19.860834Z","signature_b64":"eGno/tMhuuc07oVcyUxpbBYR6Q5BisD8vbJ9BSVjtrvhWDNshgxXHs8PEWVJDo6TS1g5auNScsZhTues/uaQDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1985061488af35b0b13024353e0c167a43775205c5aa2876e51aa725408c3448","last_reissued_at":"2026-07-05T11:06:19.860380Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:19.860380Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Prompt Perturbation in Retrieval-Augmented Generation based Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Chen Wang, Helen (Hye-Young) Paik, Liming Zhu, Yanfeng Shu, Zhibo Hu","submitted_at":"2024-02-11T12:25:41Z","abstract_excerpt":"The robustness of large language models (LLMs) becomes increasingly important as their use rapidly grows in a wide range of domains. Retrieval-Augmented Generation (RAG) is considered as a means to improve the trustworthiness of text generation from LLMs. However, how the outputs from RAG-based LLMs are affected by slightly different inputs is not well studied. In this work, we find that the insertion of even a short prefix to the prompt leads to the generation of outputs far away from factually correct answers. We systematically evaluate the effect of such prefixes on RAG by introducing a nov"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.07179","kind":"arxiv","version":3},"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/2402.07179/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":"2402.07179","created_at":"2026-07-05T11:06:19.860439+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.07179v3","created_at":"2026-07-05T11:06:19.860439+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.07179","created_at":"2026-07-05T11:06:19.860439+00:00"},{"alias_kind":"pith_short_12","alias_value":"DGCQMFEIV423","created_at":"2026-07-05T11:06:19.860439+00:00"},{"alias_kind":"pith_short_16","alias_value":"DGCQMFEIV423BMJQ","created_at":"2026-07-05T11:06:19.860439+00:00"},{"alias_kind":"pith_short_8","alias_value":"DGCQMFEI","created_at":"2026-07-05T11:06:19.860439+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.15042","citing_title":"DeRAG: Black-box Adversarial Attacks on Multiple Retrieval-Augmented Generation Applications via Prompt Injection","ref_index":23,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ","json":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ.json","graph_json":"https://pith.science/api/pith-number/DGCQMFEIV423BMJQEQ2T4DAWPJ/graph.json","events_json":"https://pith.science/api/pith-number/DGCQMFEIV423BMJQEQ2T4DAWPJ/events.json","paper":"https://pith.science/paper/DGCQMFEI"},"agent_actions":{"view_html":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ","download_json":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ.json","view_paper":"https://pith.science/paper/DGCQMFEI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.07179&json=true","fetch_graph":"https://pith.science/api/pith-number/DGCQMFEIV423BMJQEQ2T4DAWPJ/graph.json","fetch_events":"https://pith.science/api/pith-number/DGCQMFEIV423BMJQEQ2T4DAWPJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ/action/storage_attestation","attest_author":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ/action/author_attestation","sign_citation":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ/action/citation_signature","submit_replication":"https://pith.science/pith/DGCQMFEIV423BMJQEQ2T4DAWPJ/action/replication_record"}},"created_at":"2026-07-05T11:06:19.860439+00:00","updated_at":"2026-07-05T11:06:19.860439+00:00"}