{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:AEOUYY32WHR2UHE6CTT2GNNVFS","short_pith_number":"pith:AEOUYY32","schema_version":"1.0","canonical_sha256":"011d4c637ab1e3aa1c9e14e7a335b52ca5dd04218d7f38bbfe8e63a51f626003","source":{"kind":"arxiv","id":"2308.13768","version":1},"attestation_state":"computed","paper":{"title":"Adversarial Fine-Tuning of Language Models: An Iterative Optimisation Approach for the Generation and Detection of Problematic Content","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Charles O'Neill, Ioana Ciuca, Jack Miller, Thang Bui, Yuan-Sen Ting","submitted_at":"2023-08-26T05:20:58Z","abstract_excerpt":"In this paper, we tackle the emerging challenge of unintended harmful content generation in Large Language Models (LLMs) with a novel dual-stage optimisation technique using adversarial fine-tuning. Our two-pronged approach employs an adversarial model, fine-tuned to generate potentially harmful prompts, and a judge model, iteratively optimised to discern these prompts. In this adversarial cycle, the two models seek to outperform each other in the prompting phase, generating a dataset of rich examples which are then used for fine-tuning. This iterative application of prompting and fine-tuning "},"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":"2308.13768","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-08-26T05:20:58Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"22b1d8f030ad9a53715fe20008dc0e2407044b152bf828c23c42f3edd7919457","abstract_canon_sha256":"c583453f2c17077fd6181e31cdb779ea842dfc57d83d3fff03178ede32e6c5ee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:45:04.676292Z","signature_b64":"bINRFXPbaOu6WPsbgTKHJPWVWOyinyrJH9wgBhRqHmIuSMhbpvnLC8ZOeb0Vl9VKIb59akmOisvQybYoE0EEAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"011d4c637ab1e3aa1c9e14e7a335b52ca5dd04218d7f38bbfe8e63a51f626003","last_reissued_at":"2026-07-05T06:45:04.675848Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:45:04.675848Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adversarial Fine-Tuning of Language Models: An Iterative Optimisation Approach for the Generation and Detection of Problematic Content","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Charles O'Neill, Ioana Ciuca, Jack Miller, Thang Bui, Yuan-Sen Ting","submitted_at":"2023-08-26T05:20:58Z","abstract_excerpt":"In this paper, we tackle the emerging challenge of unintended harmful content generation in Large Language Models (LLMs) with a novel dual-stage optimisation technique using adversarial fine-tuning. Our two-pronged approach employs an adversarial model, fine-tuned to generate potentially harmful prompts, and a judge model, iteratively optimised to discern these prompts. In this adversarial cycle, the two models seek to outperform each other in the prompting phase, generating a dataset of rich examples which are then used for fine-tuning. This iterative application of prompting and fine-tuning "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2308.13768","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/2308.13768/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":"2308.13768","created_at":"2026-07-05T06:45:04.675914+00:00"},{"alias_kind":"arxiv_version","alias_value":"2308.13768v1","created_at":"2026-07-05T06:45:04.675914+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2308.13768","created_at":"2026-07-05T06:45:04.675914+00:00"},{"alias_kind":"pith_short_12","alias_value":"AEOUYY32WHR2","created_at":"2026-07-05T06:45:04.675914+00:00"},{"alias_kind":"pith_short_16","alias_value":"AEOUYY32WHR2UHE6","created_at":"2026-07-05T06:45:04.675914+00:00"},{"alias_kind":"pith_short_8","alias_value":"AEOUYY32","created_at":"2026-07-05T06:45:04.675914+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.24369","citing_title":"Adversarial Preference Learning for Robust LLM Alignment","ref_index":29,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS","json":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS.json","graph_json":"https://pith.science/api/pith-number/AEOUYY32WHR2UHE6CTT2GNNVFS/graph.json","events_json":"https://pith.science/api/pith-number/AEOUYY32WHR2UHE6CTT2GNNVFS/events.json","paper":"https://pith.science/paper/AEOUYY32"},"agent_actions":{"view_html":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS","download_json":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS.json","view_paper":"https://pith.science/paper/AEOUYY32","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2308.13768&json=true","fetch_graph":"https://pith.science/api/pith-number/AEOUYY32WHR2UHE6CTT2GNNVFS/graph.json","fetch_events":"https://pith.science/api/pith-number/AEOUYY32WHR2UHE6CTT2GNNVFS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS/action/storage_attestation","attest_author":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS/action/author_attestation","sign_citation":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS/action/citation_signature","submit_replication":"https://pith.science/pith/AEOUYY32WHR2UHE6CTT2GNNVFS/action/replication_record"}},"created_at":"2026-07-05T06:45:04.675914+00:00","updated_at":"2026-07-05T06:45:04.675914+00:00"}