{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:57DJGWQYYH6XTNH5IGMQGGPHOR","short_pith_number":"pith:57DJGWQY","schema_version":"1.0","canonical_sha256":"efc6935a18c1fd79b4fd41990319e7746125fb521ea4dc659a4e60d9632761f4","source":{"kind":"arxiv","id":"2508.03864","version":2},"attestation_state":"computed","paper":{"title":"Evo-MARL: Co-Evolutionary Multi-Agent Reinforcement Learning for Internalized Safety","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Dennis Wu, Han Liu, Haozheng Luo, Hong-Yu Chen, Jianshu Zhang, Manling Li, Philip S. Yu, Yiting Zhang, Yutong Zhang, Yuwei Han, Zhenyu Pan","submitted_at":"2025-08-05T19:26:55Z","abstract_excerpt":"Multi-agent systems (MAS) built on multimodal large language models exhibit strong collaboration and performance. However, their growing openness and interaction complexity pose serious risks, notably jailbreak and adversarial attacks. Existing defenses typically rely on external guard modules, such as dedicated safety agents, to handle unsafe behaviors. Unfortunately, this paradigm faces two challenges: (1) standalone agents offer limited protection, and (2) their independence leads to single-point failure-if compromised, system-wide safety collapses. Naively increasing the number of guard ag"},"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":"2508.03864","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-08-05T19:26:55Z","cross_cats_sorted":[],"title_canon_sha256":"67c9db2f2447bfa56dde0c33e3937204a858f0a308306830ee39635a7a1e51f7","abstract_canon_sha256":"bd66479f122ffd282d0f6ab8ca8f496927876885d913fff0ee85e3e341a204ae"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:05:43.881381Z","signature_b64":"41MzJWOCMZfJXlbBZO1DZhHcPL6NNNtDtuilA+gxcVpGu3vhDbk0KFdUp8i4XIOKvFWP6/nTEE6jJdjnobTkBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"efc6935a18c1fd79b4fd41990319e7746125fb521ea4dc659a4e60d9632761f4","last_reissued_at":"2026-07-05T12:05:43.880866Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:05:43.880866Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evo-MARL: Co-Evolutionary Multi-Agent Reinforcement Learning for Internalized Safety","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Dennis Wu, Han Liu, Haozheng Luo, Hong-Yu Chen, Jianshu Zhang, Manling Li, Philip S. Yu, Yiting Zhang, Yutong Zhang, Yuwei Han, Zhenyu Pan","submitted_at":"2025-08-05T19:26:55Z","abstract_excerpt":"Multi-agent systems (MAS) built on multimodal large language models exhibit strong collaboration and performance. However, their growing openness and interaction complexity pose serious risks, notably jailbreak and adversarial attacks. Existing defenses typically rely on external guard modules, such as dedicated safety agents, to handle unsafe behaviors. Unfortunately, this paradigm faces two challenges: (1) standalone agents offer limited protection, and (2) their independence leads to single-point failure-if compromised, system-wide safety collapses. Naively increasing the number of guard ag"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.03864","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/2508.03864/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":"2508.03864","created_at":"2026-07-05T12:05:43.880921+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.03864v2","created_at":"2026-07-05T12:05:43.880921+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.03864","created_at":"2026-07-05T12:05:43.880921+00:00"},{"alias_kind":"pith_short_12","alias_value":"57DJGWQYYH6X","created_at":"2026-07-05T12:05:43.880921+00:00"},{"alias_kind":"pith_short_16","alias_value":"57DJGWQYYH6XTNH5","created_at":"2026-07-05T12:05:43.880921+00:00"},{"alias_kind":"pith_short_8","alias_value":"57DJGWQY","created_at":"2026-07-05T12:05:43.880921+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2511.12710","citing_title":"Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR","json":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR.json","graph_json":"https://pith.science/api/pith-number/57DJGWQYYH6XTNH5IGMQGGPHOR/graph.json","events_json":"https://pith.science/api/pith-number/57DJGWQYYH6XTNH5IGMQGGPHOR/events.json","paper":"https://pith.science/paper/57DJGWQY"},"agent_actions":{"view_html":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR","download_json":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR.json","view_paper":"https://pith.science/paper/57DJGWQY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.03864&json=true","fetch_graph":"https://pith.science/api/pith-number/57DJGWQYYH6XTNH5IGMQGGPHOR/graph.json","fetch_events":"https://pith.science/api/pith-number/57DJGWQYYH6XTNH5IGMQGGPHOR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR/action/storage_attestation","attest_author":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR/action/author_attestation","sign_citation":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR/action/citation_signature","submit_replication":"https://pith.science/pith/57DJGWQYYH6XTNH5IGMQGGPHOR/action/replication_record"}},"created_at":"2026-07-05T12:05:43.880921+00:00","updated_at":"2026-07-05T12:05:43.880921+00:00"}