{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FXSPYSGK4PSBLI4G62QLT5VJJV","short_pith_number":"pith:FXSPYSGK","schema_version":"1.0","canonical_sha256":"2de4fc48cae3e415a386f6a0b9f6a94d73d11d1a57656bee8b92849b00656c77","source":{"kind":"arxiv","id":"2606.09701","version":1},"attestation_state":"computed","paper":{"title":"Learning to Attack and Defend: Adaptive Red Teaming of Language Models via GRPO","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Amanda Minnich, Blake Bullwinkel, Eugenia Kim, Mark Russinovich","submitted_at":"2026-06-08T16:21:36Z","abstract_excerpt":"AI red teaming must continually adapt to evolving attackers and defenders. Reinforcement learning offers a promising approach to discovering novel attacks, and co-training methods can produce more robust defenders in tandem. Recent works have demonstrated the efficacy of attacker-defender co-training by applying PPO and DPO, but report that GRPO is unstable in this setting. We introduce AdvGRPO, a co-training framework that makes GRPO viable for joint attacker-defender optimization using dense multi-channel rewards and decoupled advantage normalization. Training progresses through a curriculum"},"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":"2606.09701","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-08T16:21:36Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"335605a47ccb43370886ca71509a5347e0b22bb3be3413fc87ed97d1b1a15a64","abstract_canon_sha256":"a5d6e8c5ac05ddedb6f684fb2ef5e57bf138762b969913d36d8758b910be7e69"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:09:04.671779Z","signature_b64":"gGnuOU7csLpiqmtA6DTR8E0Rt3vBpxJlwlC1DdTZBE+lr4it2Myv6E8JkO2UcNvbxCL84eLNmi+nzZAyBV1fDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2de4fc48cae3e415a386f6a0b9f6a94d73d11d1a57656bee8b92849b00656c77","last_reissued_at":"2026-06-09T02:09:04.664889Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:09:04.664889Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to Attack and Defend: Adaptive Red Teaming of Language Models via GRPO","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Amanda Minnich, Blake Bullwinkel, Eugenia Kim, Mark Russinovich","submitted_at":"2026-06-08T16:21:36Z","abstract_excerpt":"AI red teaming must continually adapt to evolving attackers and defenders. Reinforcement learning offers a promising approach to discovering novel attacks, and co-training methods can produce more robust defenders in tandem. Recent works have demonstrated the efficacy of attacker-defender co-training by applying PPO and DPO, but report that GRPO is unstable in this setting. We introduce AdvGRPO, a co-training framework that makes GRPO viable for joint attacker-defender optimization using dense multi-channel rewards and decoupled advantage normalization. Training progresses through a curriculum"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.09701","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/2606.09701/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":"2606.09701","created_at":"2026-06-09T02:09:04.669535+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.09701v1","created_at":"2026-06-09T02:09:04.669535+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.09701","created_at":"2026-06-09T02:09:04.669535+00:00"},{"alias_kind":"pith_short_12","alias_value":"FXSPYSGK4PSB","created_at":"2026-06-09T02:09:04.669535+00:00"},{"alias_kind":"pith_short_16","alias_value":"FXSPYSGK4PSBLI4G","created_at":"2026-06-09T02:09:04.669535+00:00"},{"alias_kind":"pith_short_8","alias_value":"FXSPYSGK","created_at":"2026-06-09T02:09:04.669535+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/FXSPYSGK4PSBLI4G62QLT5VJJV","json":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV.json","graph_json":"https://pith.science/api/pith-number/FXSPYSGK4PSBLI4G62QLT5VJJV/graph.json","events_json":"https://pith.science/api/pith-number/FXSPYSGK4PSBLI4G62QLT5VJJV/events.json","paper":"https://pith.science/paper/FXSPYSGK"},"agent_actions":{"view_html":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV","download_json":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV.json","view_paper":"https://pith.science/paper/FXSPYSGK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.09701&json=true","fetch_graph":"https://pith.science/api/pith-number/FXSPYSGK4PSBLI4G62QLT5VJJV/graph.json","fetch_events":"https://pith.science/api/pith-number/FXSPYSGK4PSBLI4G62QLT5VJJV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV/action/storage_attestation","attest_author":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV/action/author_attestation","sign_citation":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV/action/citation_signature","submit_replication":"https://pith.science/pith/FXSPYSGK4PSBLI4G62QLT5VJJV/action/replication_record"}},"created_at":"2026-06-09T02:09:04.669535+00:00","updated_at":"2026-06-09T02:09:04.669535+00:00"}