{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:7IXEV34H76OL7BKAPEPINWBM3P","short_pith_number":"pith:7IXEV34H","schema_version":"1.0","canonical_sha256":"fa2e4aef87ff9cbf8540791e86d82cdbedd77c59e9f2bed3bfd884f4d86e7f97","source":{"kind":"arxiv","id":"2401.05998","version":1},"attestation_state":"computed","paper":{"title":"Combating Adversarial Attacks with Multi-Agent Debate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Andy Liu, Steffi Chern, Zhen Fan","submitted_at":"2024-01-11T15:57:38Z","abstract_excerpt":"While state-of-the-art language models have achieved impressive results, they remain susceptible to inference-time adversarial attacks, such as adversarial prompts generated by red teams arXiv:2209.07858. One approach proposed to improve the general quality of language model generations is multi-agent debate, where language models self-evaluate through discussion and feedback arXiv:2305.14325. We implement multi-agent debate between current state-of-the-art language models and evaluate models' susceptibility to red team attacks in both single- and multi-agent settings. We find that multi-agent"},"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":"2401.05998","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-01-11T15:57:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1cc7998510c5cba40835a007f9c7ff226581516fa0dc72efd347f95c74c8b40c","abstract_canon_sha256":"8f357b39d5c876d99fc60bd50eeb96480b1eacc3b2851786bd36e1423fd5b52e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:32:39.302362Z","signature_b64":"tTT/LBTw+ISfBDTj3pK4W8jzvmWESSHqq1kdiwyZlMrcRnKbXz3VGclE9yBBCLoHH+qYdnhxpLiAaDWyzlhSBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fa2e4aef87ff9cbf8540791e86d82cdbedd77c59e9f2bed3bfd884f4d86e7f97","last_reissued_at":"2026-07-05T07:32:39.301946Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:32:39.301946Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Combating Adversarial Attacks with Multi-Agent Debate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Andy Liu, Steffi Chern, Zhen Fan","submitted_at":"2024-01-11T15:57:38Z","abstract_excerpt":"While state-of-the-art language models have achieved impressive results, they remain susceptible to inference-time adversarial attacks, such as adversarial prompts generated by red teams arXiv:2209.07858. One approach proposed to improve the general quality of language model generations is multi-agent debate, where language models self-evaluate through discussion and feedback arXiv:2305.14325. We implement multi-agent debate between current state-of-the-art language models and evaluate models' susceptibility to red team attacks in both single- and multi-agent settings. We find that multi-agent"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.05998","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/2401.05998/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":"2401.05998","created_at":"2026-07-05T07:32:39.302008+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.05998v1","created_at":"2026-07-05T07:32:39.302008+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.05998","created_at":"2026-07-05T07:32:39.302008+00:00"},{"alias_kind":"pith_short_12","alias_value":"7IXEV34H76OL","created_at":"2026-07-05T07:32:39.302008+00:00"},{"alias_kind":"pith_short_16","alias_value":"7IXEV34H76OL7BKA","created_at":"2026-07-05T07:32:39.302008+00:00"},{"alias_kind":"pith_short_8","alias_value":"7IXEV34H","created_at":"2026-07-05T07:32:39.302008+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19826","citing_title":"Heterogeneous LLM Debate Under Adversarial Peers: Honest Gains, Replacement Costs, and Resilience","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2503.21460","citing_title":"Large Language Model Agent: A Survey on Methodology, Applications and Challenges","ref_index":183,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14564","citing_title":"MARS$^2$: Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P","json":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P.json","graph_json":"https://pith.science/api/pith-number/7IXEV34H76OL7BKAPEPINWBM3P/graph.json","events_json":"https://pith.science/api/pith-number/7IXEV34H76OL7BKAPEPINWBM3P/events.json","paper":"https://pith.science/paper/7IXEV34H"},"agent_actions":{"view_html":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P","download_json":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P.json","view_paper":"https://pith.science/paper/7IXEV34H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.05998&json=true","fetch_graph":"https://pith.science/api/pith-number/7IXEV34H76OL7BKAPEPINWBM3P/graph.json","fetch_events":"https://pith.science/api/pith-number/7IXEV34H76OL7BKAPEPINWBM3P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P/action/storage_attestation","attest_author":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P/action/author_attestation","sign_citation":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P/action/citation_signature","submit_replication":"https://pith.science/pith/7IXEV34H76OL7BKAPEPINWBM3P/action/replication_record"}},"created_at":"2026-07-05T07:32:39.302008+00:00","updated_at":"2026-07-05T07:32:39.302008+00:00"}