{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:Y7IECXPCM7XKLXKNBSO6EC2C3B","short_pith_number":"pith:Y7IECXPC","schema_version":"1.0","canonical_sha256":"c7d0415de267eea5dd4d0c9de20b42d865f1980d1db797b345658e8fd68b5ce6","source":{"kind":"arxiv","id":"2403.17710","version":5},"attestation_state":"computed","paper":{"title":"Optimization-based Prompt Injection Attack to LLM-as-a-Judge","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Jiawen Shi, Lichao Sun, Neil Zhenqiang Gong, Pan Zhou, Yinuo Liu, Yue Huang, Zenghui Yuan","submitted_at":"2024-03-26T13:58:00Z","abstract_excerpt":"LLM-as-a-Judge uses a large language model (LLM) to select the best response from a set of candidates for a given question. LLM-as-a-Judge has many applications such as LLM-powered search, reinforcement learning with AI feedback (RLAIF), and tool selection. In this work, we propose JudgeDeceiver, an optimization-based prompt injection attack to LLM-as-a-Judge. JudgeDeceiver injects a carefully crafted sequence into an attacker-controlled candidate response such that LLM-as-a-Judge selects the candidate response for an attacker-chosen question no matter what other candidate responses are. Speci"},"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":"2403.17710","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-03-26T13:58:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"237429e534237dc56aeefb604f8338060774394b3b1936a2aface0e95d9425e9","abstract_canon_sha256":"401c236df245f1140129732e57fee63acf52912af49d70b7d92ec35e0c0b3c93"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:58:18.010820Z","signature_b64":"fk5tdL7rtxbda130n53H1M31wqMAchpCKxC8EVb5swc7acBPT6NX6mvmYqdvvah7uCTrnTnJmDs5Zdk6H9OtAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c7d0415de267eea5dd4d0c9de20b42d865f1980d1db797b345658e8fd68b5ce6","last_reissued_at":"2026-07-05T11:58:18.010310Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:58:18.010310Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimization-based Prompt Injection Attack to LLM-as-a-Judge","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Jiawen Shi, Lichao Sun, Neil Zhenqiang Gong, Pan Zhou, Yinuo Liu, Yue Huang, Zenghui Yuan","submitted_at":"2024-03-26T13:58:00Z","abstract_excerpt":"LLM-as-a-Judge uses a large language model (LLM) to select the best response from a set of candidates for a given question. LLM-as-a-Judge has many applications such as LLM-powered search, reinforcement learning with AI feedback (RLAIF), and tool selection. In this work, we propose JudgeDeceiver, an optimization-based prompt injection attack to LLM-as-a-Judge. JudgeDeceiver injects a carefully crafted sequence into an attacker-controlled candidate response such that LLM-as-a-Judge selects the candidate response for an attacker-chosen question no matter what other candidate responses are. Speci"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.17710","kind":"arxiv","version":5},"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/2403.17710/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":"2403.17710","created_at":"2026-07-05T11:58:18.010383+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.17710v5","created_at":"2026-07-05T11:58:18.010383+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.17710","created_at":"2026-07-05T11:58:18.010383+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y7IECXPCM7XK","created_at":"2026-07-05T11:58:18.010383+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y7IECXPCM7XKLXKN","created_at":"2026-07-05T11:58:18.010383+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y7IECXPC","created_at":"2026-07-05T11:58:18.010383+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2504.20472","citing_title":"Robustness via Referencing: Defending against Prompt Injection Attacks by Referencing the Executed Instruction","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2505.19237","citing_title":"Sensorimotor Self-Recognition in Multimodal Large Language Model-Driven Robots","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2510.09093","citing_title":"Exploiting Web Search Tools of AI Agents for Data Exfiltration","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2412.05579","citing_title":"LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods","ref_index":203,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10134","citing_title":"PlanGuard: Defending Agents against Indirect Prompt Injection via Planning-based Consistency Verification","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13602","citing_title":"Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B","json":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B.json","graph_json":"https://pith.science/api/pith-number/Y7IECXPCM7XKLXKNBSO6EC2C3B/graph.json","events_json":"https://pith.science/api/pith-number/Y7IECXPCM7XKLXKNBSO6EC2C3B/events.json","paper":"https://pith.science/paper/Y7IECXPC"},"agent_actions":{"view_html":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B","download_json":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B.json","view_paper":"https://pith.science/paper/Y7IECXPC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.17710&json=true","fetch_graph":"https://pith.science/api/pith-number/Y7IECXPCM7XKLXKNBSO6EC2C3B/graph.json","fetch_events":"https://pith.science/api/pith-number/Y7IECXPCM7XKLXKNBSO6EC2C3B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B/action/storage_attestation","attest_author":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B/action/author_attestation","sign_citation":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B/action/citation_signature","submit_replication":"https://pith.science/pith/Y7IECXPCM7XKLXKNBSO6EC2C3B/action/replication_record"}},"created_at":"2026-07-05T11:58:18.010383+00:00","updated_at":"2026-07-05T11:58:18.010383+00:00"}