{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CSGQQKG32NFXOPUTI2XQLHW5FL","short_pith_number":"pith:CSGQQKG3","schema_version":"1.0","canonical_sha256":"148d0828dbd34b773e9346af059edd2ae8bd9b7a6dc40b3d6cd02e01c0b2b854","source":{"kind":"arxiv","id":"2405.15589","version":3},"attestation_state":"computed","paper":{"title":"Efficient Adversarial Training in LLMs with Continuous Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Alessandro Sordoni, Gauthier Gidel, Leo Schwinn, Sophie Xhonneux, Stephan G\\\"unnemann","submitted_at":"2024-05-24T14:20:09Z","abstract_excerpt":"Large language models (LLMs) are vulnerable to adversarial attacks that can bypass their safety guardrails. In many domains, adversarial training has proven to be one of the most promising methods to reliably improve robustness against such attacks. Yet, in the context of LLMs, current methods for adversarial training are hindered by the high computational costs required to perform discrete adversarial attacks at each training iteration. We address this problem by instead calculating adversarial attacks in the continuous embedding space of the LLM, which is orders of magnitudes more efficient."},"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":"2405.15589","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-24T14:20:09Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"ba0b9bf395da460bdb869dc943d7a7fe8a5a029a92ee806781cbeaf0ca3da718","abstract_canon_sha256":"caec8f9c347aa298f45f4e659879ef107fa55158635765f88cd8e6d5a55e88b0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:29:28.987959Z","signature_b64":"3wH4oUIgZPTfW7ZkbZK3+H+3+U/x2FkaX/d4fM89N6B9yi7nsbUefSHLx2O0L/t9QCeeG9iYwEnLkaO8tpEIBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"148d0828dbd34b773e9346af059edd2ae8bd9b7a6dc40b3d6cd02e01c0b2b854","last_reissued_at":"2026-07-05T09:29:28.987479Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:29:28.987479Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Adversarial Training in LLMs with Continuous Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Alessandro Sordoni, Gauthier Gidel, Leo Schwinn, Sophie Xhonneux, Stephan G\\\"unnemann","submitted_at":"2024-05-24T14:20:09Z","abstract_excerpt":"Large language models (LLMs) are vulnerable to adversarial attacks that can bypass their safety guardrails. In many domains, adversarial training has proven to be one of the most promising methods to reliably improve robustness against such attacks. Yet, in the context of LLMs, current methods for adversarial training are hindered by the high computational costs required to perform discrete adversarial attacks at each training iteration. We address this problem by instead calculating adversarial attacks in the continuous embedding space of the LLM, which is orders of magnitudes more efficient."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.15589","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/2405.15589/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":"2405.15589","created_at":"2026-07-05T09:29:28.987538+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.15589v3","created_at":"2026-07-05T09:29:28.987538+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.15589","created_at":"2026-07-05T09:29:28.987538+00:00"},{"alias_kind":"pith_short_12","alias_value":"CSGQQKG32NFX","created_at":"2026-07-05T09:29:28.987538+00:00"},{"alias_kind":"pith_short_16","alias_value":"CSGQQKG32NFXOPUT","created_at":"2026-07-05T09:29:28.987538+00:00"},{"alias_kind":"pith_short_8","alias_value":"CSGQQKG3","created_at":"2026-07-05T09:29:28.987538+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07918","citing_title":"Efficient Safety Alignment of Language Models via Latent Personality Traits","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2606.05523","citing_title":"CHASE: Adversarial Red-Blue Teaming for Improving LLM Safety using Reinforcement Learning","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29816","citing_title":"Harnessing non-adversarial robustness in large language models","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20519","citing_title":"Codec-Robust Attacks on Audio LLMs","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2503.02574","citing_title":"LLM-Safety Evaluations Lack Robustness","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2505.16120","citing_title":"LLM-Powered AI Agent Systems and Their Applications in Industry","ref_index":118,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20519","citing_title":"Codec-Robust Attacks on Audio LLMs","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12813","citing_title":"REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11170","citing_title":"Unlearning with Asymmetric Sources: Improved Unlearning-Utility Trade-off with Public Data","ref_index":130,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL","json":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL.json","graph_json":"https://pith.science/api/pith-number/CSGQQKG32NFXOPUTI2XQLHW5FL/graph.json","events_json":"https://pith.science/api/pith-number/CSGQQKG32NFXOPUTI2XQLHW5FL/events.json","paper":"https://pith.science/paper/CSGQQKG3"},"agent_actions":{"view_html":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL","download_json":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL.json","view_paper":"https://pith.science/paper/CSGQQKG3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.15589&json=true","fetch_graph":"https://pith.science/api/pith-number/CSGQQKG32NFXOPUTI2XQLHW5FL/graph.json","fetch_events":"https://pith.science/api/pith-number/CSGQQKG32NFXOPUTI2XQLHW5FL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL/action/storage_attestation","attest_author":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL/action/author_attestation","sign_citation":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL/action/citation_signature","submit_replication":"https://pith.science/pith/CSGQQKG32NFXOPUTI2XQLHW5FL/action/replication_record"}},"created_at":"2026-07-05T09:29:28.987538+00:00","updated_at":"2026-07-05T09:29:28.987538+00:00"}