{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:3EUUJX67Z5S2ATHNLCN7UPXCJA","short_pith_number":"pith:3EUUJX67","canonical_record":{"source":{"id":"2604.27019","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-29T12:44:05Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"927b03b1fcd80ef19340f467342c19c575ccadfebac359d713a90831292e5fed","abstract_canon_sha256":"a33e5c77f8972c6899dffa859d630a2f0e07185da66e708ef465b7cc16a55450"},"schema_version":"1.0"},"canonical_sha256":"d92944dfdfcf65a04ced589bfa3ee2481bd135a94f0d604ef1c05e5bafefb3ca","source":{"kind":"arxiv","id":"2604.27019","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.27019","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"arxiv_version","alias_value":"2604.27019v2","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.27019","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_12","alias_value":"3EUUJX67Z5S2","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_16","alias_value":"3EUUJX67Z5S2ATHN","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_8","alias_value":"3EUUJX67","created_at":"2026-05-20T00:03:13Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:3EUUJX67Z5S2ATHNLCN7UPXCJA","target":"record","payload":{"canonical_record":{"source":{"id":"2604.27019","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-29T12:44:05Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"927b03b1fcd80ef19340f467342c19c575ccadfebac359d713a90831292e5fed","abstract_canon_sha256":"a33e5c77f8972c6899dffa859d630a2f0e07185da66e708ef465b7cc16a55450"},"schema_version":"1.0"},"canonical_sha256":"d92944dfdfcf65a04ced589bfa3ee2481bd135a94f0d604ef1c05e5bafefb3ca","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:13.065351Z","signature_b64":"uhozzIjoPL0pU7A3a6rNsQWtMgqYXTiG5R78hl0nVzjx7DWAdxBGvaYu5vmKTl0rSL/et0akBoVOxBb3kZ6IBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d92944dfdfcf65a04ced589bfa3ee2481bd135a94f0d604ef1c05e5bafefb3ca","last_reissued_at":"2026-05-20T00:03:13.064371Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:13.064371Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.27019","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mhiwnITRJdhP5atSsllZPkYNHaRzQWhjXhBMqSYq31VD6FgTYIs+XtyLdPAfRkwlqyqqxnL0WwCXSSogT4rkAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-24T23:54:38.972048Z"},"content_sha256":"e2ad78177acee8511a021ca690e0ec46d2541ce747c52e6505c01f03640b7d21","schema_version":"1.0","event_id":"sha256:e2ad78177acee8511a021ca690e0ec46d2541ce747c52e6505c01f03640b7d21"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:3EUUJX67Z5S2ATHNLCN7UPXCJA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Dynamic Adversarial Fine-Tuning Reorganizes Refusal Geometry","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Dynamic adversarial fine-tuning reorganizes refusal geometry by relocating the primary carrier from late layers to early layers.","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Haihua Shen, Junbin Yang, Shan Li, Wenhao Lan, Yijun Yang","submitted_at":"2026-04-29T12:44:05Z","abstract_excerpt":"Safety-aligned language models must refuse harmful requests without collapsing into broad over-refusal, yet it remains unclear how dynamic adversarial fine-tuning changes the internal carriers of refusal. We study one 7B backbone under supervised fine-tuning (SFT) and under Robust Refusal Dynamic Defense (R2D2), a HarmBench-style adversarial fine-tuning procedure that repeatedly refreshes harmful training cases with current jailbreak attacks. Our protocol aligns fixed-source HarmBench, StrongREJECT, and XSTest with a five-anchor refusal-geometry suite, causal interventions, and a sparse adapti"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"These results support a reorganization account rather than a drift-only account, with evidence limited to one backbone and fixed-source attacks. R2D2 preserves a late-layer admissible carrier through step 100 before relocating to an early-layer carrier, while effective rank remains near 1.23--1.27.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the five-anchor refusal-geometry suite and causal interventions accurately isolate reorganization effects without being confounded by the specific 7B backbone, fixed attack sources, or unmeasured factors in the training dynamics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"R2D2-style dynamic adversarial fine-tuning reorganizes refusal geometry from late-layer to early-layer carriers in LLMs, achieving lower attack success rates than SFT while maintaining low-dimensional structure.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Dynamic adversarial fine-tuning reorganizes refusal geometry by relocating the primary carrier from late layers to early layers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"265c253502e11c0992554407c8dbb6ac506b4fd15e594aafd951c423a0df48b5"},"source":{"id":"2604.27019","kind":"arxiv","version":2},"verdict":{"id":"93cead99-4432-44c0-89a2-9ac4e8473454","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T11:32:46.163951Z","strongest_claim":"These results support a reorganization account rather than a drift-only account, with evidence limited to one backbone and fixed-source attacks. R2D2 preserves a late-layer admissible carrier through step 100 before relocating to an early-layer carrier, while effective rank remains near 1.23--1.27.","one_line_summary":"R2D2-style dynamic adversarial fine-tuning reorganizes refusal geometry from late-layer to early-layer carriers in LLMs, achieving lower attack success rates than SFT while maintaining low-dimensional structure.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the five-anchor refusal-geometry suite and causal interventions accurately isolate reorganization effects without being confounded by the specific 7B backbone, fixed attack sources, or unmeasured factors in the training dynamics.","pith_extraction_headline":"Dynamic adversarial fine-tuning reorganizes refusal geometry by relocating the primary carrier from late layers to early layers."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.27019/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T19:58:34.044956Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"5847dc47980c3abd0858840420a67c44eebbfd69949c723ee090881a0ba9d545"},"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"},"verdict_id":"93cead99-4432-44c0-89a2-9ac4e8473454"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Dl29DVbTitkqFDQk5HRDX7yULzqAwVkPitSUKENpLZ+Y2YAQlIelGUtY7MSgT60SRB+TgTre3Nc3/YCg0w5lBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-24T23:54:38.972892Z"},"content_sha256":"63c175f434c6bc04b3ceea637c4dbe1b008ab4b8f61f99f9db07d80eb39de3fc","schema_version":"1.0","event_id":"sha256:63c175f434c6bc04b3ceea637c4dbe1b008ab4b8f61f99f9db07d80eb39de3fc"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA/bundle.json","state_url":"https://pith.science/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-24T23:54:38Z","links":{"resolver":"https://pith.science/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA","bundle":"https://pith.science/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA/bundle.json","state":"https://pith.science/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3EUUJX67Z5S2ATHNLCN7UPXCJA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:3EUUJX67Z5S2ATHNLCN7UPXCJA","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"a33e5c77f8972c6899dffa859d630a2f0e07185da66e708ef465b7cc16a55450","cross_cats_sorted":["cs.CL","cs.CR"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-29T12:44:05Z","title_canon_sha256":"927b03b1fcd80ef19340f467342c19c575ccadfebac359d713a90831292e5fed"},"schema_version":"1.0","source":{"id":"2604.27019","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.27019","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"arxiv_version","alias_value":"2604.27019v2","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.27019","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_12","alias_value":"3EUUJX67Z5S2","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_16","alias_value":"3EUUJX67Z5S2ATHN","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_8","alias_value":"3EUUJX67","created_at":"2026-05-20T00:03:13Z"}],"graph_snapshots":[{"event_id":"sha256:63c175f434c6bc04b3ceea637c4dbe1b008ab4b8f61f99f9db07d80eb39de3fc","target":"graph","created_at":"2026-05-20T00:03:13Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"These results support a reorganization account rather than a drift-only account, with evidence limited to one backbone and fixed-source attacks. R2D2 preserves a late-layer admissible carrier through step 100 before relocating to an early-layer carrier, while effective rank remains near 1.23--1.27."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the five-anchor refusal-geometry suite and causal interventions accurately isolate reorganization effects without being confounded by the specific 7B backbone, fixed attack sources, or unmeasured factors in the training dynamics."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"R2D2-style dynamic adversarial fine-tuning reorganizes refusal geometry from late-layer to early-layer carriers in LLMs, achieving lower attack success rates than SFT while maintaining low-dimensional structure."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Dynamic adversarial fine-tuning reorganizes refusal geometry by relocating the primary carrier from late layers to early layers."}],"snapshot_sha256":"265c253502e11c0992554407c8dbb6ac506b4fd15e594aafd951c423a0df48b5"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T19:58:34.044956Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.27019/integrity.json","findings":[],"snapshot_sha256":"5847dc47980c3abd0858840420a67c44eebbfd69949c723ee090881a0ba9d545","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Safety-aligned language models must refuse harmful requests without collapsing into broad over-refusal, yet it remains unclear how dynamic adversarial fine-tuning changes the internal carriers of refusal. We study one 7B backbone under supervised fine-tuning (SFT) and under Robust Refusal Dynamic Defense (R2D2), a HarmBench-style adversarial fine-tuning procedure that repeatedly refreshes harmful training cases with current jailbreak attacks. Our protocol aligns fixed-source HarmBench, StrongREJECT, and XSTest with a five-anchor refusal-geometry suite, causal interventions, and a sparse adapti","authors_text":"Haihua Shen, Junbin Yang, Shan Li, Wenhao Lan, Yijun Yang","cross_cats":["cs.CL","cs.CR"],"headline":"Dynamic adversarial fine-tuning reorganizes refusal geometry by relocating the primary carrier from late layers to early layers.","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-29T12:44:05Z","title":"Dynamic Adversarial Fine-Tuning Reorganizes Refusal Geometry"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.27019","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-07T11:32:46.163951Z","id":"93cead99-4432-44c0-89a2-9ac4e8473454","model_set":{"reader":"grok-4.3"},"one_line_summary":"R2D2-style dynamic adversarial fine-tuning reorganizes refusal geometry from late-layer to early-layer carriers in LLMs, achieving lower attack success rates than SFT while maintaining low-dimensional structure.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Dynamic adversarial fine-tuning reorganizes refusal geometry by relocating the primary carrier from late layers to early layers.","strongest_claim":"These results support a reorganization account rather than a drift-only account, with evidence limited to one backbone and fixed-source attacks. R2D2 preserves a late-layer admissible carrier through step 100 before relocating to an early-layer carrier, while effective rank remains near 1.23--1.27.","weakest_assumption":"That the five-anchor refusal-geometry suite and causal interventions accurately isolate reorganization effects without being confounded by the specific 7B backbone, fixed attack sources, or unmeasured factors in the training dynamics."}},"verdict_id":"93cead99-4432-44c0-89a2-9ac4e8473454"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e2ad78177acee8511a021ca690e0ec46d2541ce747c52e6505c01f03640b7d21","target":"record","created_at":"2026-05-20T00:03:13Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"a33e5c77f8972c6899dffa859d630a2f0e07185da66e708ef465b7cc16a55450","cross_cats_sorted":["cs.CL","cs.CR"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-29T12:44:05Z","title_canon_sha256":"927b03b1fcd80ef19340f467342c19c575ccadfebac359d713a90831292e5fed"},"schema_version":"1.0","source":{"id":"2604.27019","kind":"arxiv","version":2}},"canonical_sha256":"d92944dfdfcf65a04ced589bfa3ee2481bd135a94f0d604ef1c05e5bafefb3ca","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d92944dfdfcf65a04ced589bfa3ee2481bd135a94f0d604ef1c05e5bafefb3ca","first_computed_at":"2026-05-20T00:03:13.064371Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:13.064371Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"uhozzIjoPL0pU7A3a6rNsQWtMgqYXTiG5R78hl0nVzjx7DWAdxBGvaYu5vmKTl0rSL/et0akBoVOxBb3kZ6IBQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:13.065351Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.27019","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e2ad78177acee8511a021ca690e0ec46d2541ce747c52e6505c01f03640b7d21","sha256:63c175f434c6bc04b3ceea637c4dbe1b008ab4b8f61f99f9db07d80eb39de3fc"],"state_sha256":"1905c74737a3246309432b7a035f58d4f688647843726006be444b6db8a9ba69"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"O5FGdF5ZHI60ldcyneFK/JjxduH0zrYel58/Co6OY9vnT1S8YDJdbIKrS1IP7rjZS2VU4FZzXlQiqmIPe4qHBA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-24T23:54:38.976750Z","bundle_sha256":"90b15ede0cab3d8c0a890672159e76522e81f5f6f3d67fb0fc088a917ecc0323"}}