{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2FMEWEV4LTKGZNCUE4KVFYZVXL","short_pith_number":"pith:2FMEWEV4","schema_version":"1.0","canonical_sha256":"d1584b12bc5cd46cb454271552e335bae34544cb430929a632b0a09bce284576","source":{"kind":"arxiv","id":"2605.31140","version":1},"attestation_state":"computed","paper":{"title":"EvoDefense: Co-Evolving Black-Box Defense with Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Chaochao Lu, Yanming Guo, Yingmei Wei, Yuenan Hou, Yu Li","submitted_at":"2026-05-29T10:49:44Z","abstract_excerpt":"Large Language Models (LLMs) remain highly vulnerable to diverse attacks, particularly in black-box settings where the internals of target models are inaccessible. Existing black-box defenses typically rely on pre-defined filtering heuristics, which often fail to generalize to unseen attack types and target model architectures. We introduce EvoDefense, an experience-guided co-evolving black-box defense paradigm. EvoDefense employs a guard LLM to detect malicious queries and an experience memory module to accumulate defense knowledge from previous interactions. At the core of EvoDefense is a co"},"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":"2605.31140","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-05-29T10:49:44Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"f0bfe3a23c2bc7ce9f35081219eea6bba35ae12548f933a2a9c807dcb717fdd9","abstract_canon_sha256":"6a53e06217d3e538d031043a62edde3d3c718ceca7d4fc8bbc22d0ed30a8e4a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:04:00.724680Z","signature_b64":"3wLq91OuOa4SjzaQJMCQG30CWcUOnxtbNIGRLhhYypSHpGcJYdGHBpx2c3sXVN7xGBEDuzq7ofWSgjOcr5hmAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d1584b12bc5cd46cb454271552e335bae34544cb430929a632b0a09bce284576","last_reissued_at":"2026-06-01T01:04:00.723698Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:04:00.723698Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EvoDefense: Co-Evolving Black-Box Defense with Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Chaochao Lu, Yanming Guo, Yingmei Wei, Yuenan Hou, Yu Li","submitted_at":"2026-05-29T10:49:44Z","abstract_excerpt":"Large Language Models (LLMs) remain highly vulnerable to diverse attacks, particularly in black-box settings where the internals of target models are inaccessible. Existing black-box defenses typically rely on pre-defined filtering heuristics, which often fail to generalize to unseen attack types and target model architectures. We introduce EvoDefense, an experience-guided co-evolving black-box defense paradigm. EvoDefense employs a guard LLM to detect malicious queries and an experience memory module to accumulate defense knowledge from previous interactions. At the core of EvoDefense is a co"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.31140","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/2605.31140/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":"2605.31140","created_at":"2026-06-01T01:04:00.723892+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.31140v1","created_at":"2026-06-01T01:04:00.723892+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.31140","created_at":"2026-06-01T01:04:00.723892+00:00"},{"alias_kind":"pith_short_12","alias_value":"2FMEWEV4LTKG","created_at":"2026-06-01T01:04:00.723892+00:00"},{"alias_kind":"pith_short_16","alias_value":"2FMEWEV4LTKGZNCU","created_at":"2026-06-01T01:04:00.723892+00:00"},{"alias_kind":"pith_short_8","alias_value":"2FMEWEV4","created_at":"2026-06-01T01:04:00.723892+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL","json":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL.json","graph_json":"https://pith.science/api/pith-number/2FMEWEV4LTKGZNCUE4KVFYZVXL/graph.json","events_json":"https://pith.science/api/pith-number/2FMEWEV4LTKGZNCUE4KVFYZVXL/events.json","paper":"https://pith.science/paper/2FMEWEV4"},"agent_actions":{"view_html":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL","download_json":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL.json","view_paper":"https://pith.science/paper/2FMEWEV4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.31140&json=true","fetch_graph":"https://pith.science/api/pith-number/2FMEWEV4LTKGZNCUE4KVFYZVXL/graph.json","fetch_events":"https://pith.science/api/pith-number/2FMEWEV4LTKGZNCUE4KVFYZVXL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL/action/storage_attestation","attest_author":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL/action/author_attestation","sign_citation":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL/action/citation_signature","submit_replication":"https://pith.science/pith/2FMEWEV4LTKGZNCUE4KVFYZVXL/action/replication_record"}},"created_at":"2026-06-01T01:04:00.723892+00:00","updated_at":"2026-06-01T01:04:00.723892+00:00"}