{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GNAYXLFD3NAFAYUBOM33MFYGE7","short_pith_number":"pith:GNAYXLFD","schema_version":"1.0","canonical_sha256":"33418baca3db405062817337b6170627c545b7cef71fb7ea5c3b769dc6d78707","source":{"kind":"arxiv","id":"2606.22676","version":1},"attestation_state":"computed","paper":{"title":"Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Dongyub Jude Lee, Heuiseok Lim, Jaehyung Seo, Jungseob Lee, Seongtae Hong, Seungyoon Lee, Sugyeong Eo, Suhyune Son","submitted_at":"2026-06-21T21:30:15Z","abstract_excerpt":"Alignment tuning is meant to make harmful-request refusal robust, yet this safety behavior can be erased by a small set of benign fine-tuning examples. This is a deployment risk for open-weight models because a checkpoint can pass refusal tests at release time and later lose refusal under low-cost downstream fine-tuning. Prior work has established these refusal failures, but existing studies do not show how to detect this fragility in the aligned model itself before an attack or fine-tuning intervention is run. We introduce Skin-Deep, a geometric diagnostic that detects alignment fragility dir"},"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":"2606.22676","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-06-21T21:30:15Z","cross_cats_sorted":[],"title_canon_sha256":"8cea4c5779240e440827336bf282c32529756ccb4b9ef3b67ff0055c000a7ec9","abstract_canon_sha256":"9fbfde3f3d41182a9f88a780d6a7b06c397c2b813fd90d9d2bde33d9fda69e98"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T02:13:44.569594Z","signature_b64":"gkfygtJRtTuEdadOCatz6OYAZxHqzhScF5kCBjpSvqytG3MmtaeDAA3PkUKIVLRy/MDcGfEA2VVCMKjAqBuqDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"33418baca3db405062817337b6170627c545b7cef71fb7ea5c3b769dc6d78707","last_reissued_at":"2026-06-23T02:13:44.569178Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T02:13:44.569178Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Dongyub Jude Lee, Heuiseok Lim, Jaehyung Seo, Jungseob Lee, Seongtae Hong, Seungyoon Lee, Sugyeong Eo, Suhyune Son","submitted_at":"2026-06-21T21:30:15Z","abstract_excerpt":"Alignment tuning is meant to make harmful-request refusal robust, yet this safety behavior can be erased by a small set of benign fine-tuning examples. This is a deployment risk for open-weight models because a checkpoint can pass refusal tests at release time and later lose refusal under low-cost downstream fine-tuning. Prior work has established these refusal failures, but existing studies do not show how to detect this fragility in the aligned model itself before an attack or fine-tuning intervention is run. We introduce Skin-Deep, a geometric diagnostic that detects alignment fragility dir"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.22676","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/2606.22676/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":"2606.22676","created_at":"2026-06-23T02:13:44.569239+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.22676v1","created_at":"2026-06-23T02:13:44.569239+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.22676","created_at":"2026-06-23T02:13:44.569239+00:00"},{"alias_kind":"pith_short_12","alias_value":"GNAYXLFD3NAF","created_at":"2026-06-23T02:13:44.569239+00:00"},{"alias_kind":"pith_short_16","alias_value":"GNAYXLFD3NAFAYUB","created_at":"2026-06-23T02:13:44.569239+00:00"},{"alias_kind":"pith_short_8","alias_value":"GNAYXLFD","created_at":"2026-06-23T02:13:44.569239+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/GNAYXLFD3NAFAYUBOM33MFYGE7","json":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7.json","graph_json":"https://pith.science/api/pith-number/GNAYXLFD3NAFAYUBOM33MFYGE7/graph.json","events_json":"https://pith.science/api/pith-number/GNAYXLFD3NAFAYUBOM33MFYGE7/events.json","paper":"https://pith.science/paper/GNAYXLFD"},"agent_actions":{"view_html":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7","download_json":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7.json","view_paper":"https://pith.science/paper/GNAYXLFD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.22676&json=true","fetch_graph":"https://pith.science/api/pith-number/GNAYXLFD3NAFAYUBOM33MFYGE7/graph.json","fetch_events":"https://pith.science/api/pith-number/GNAYXLFD3NAFAYUBOM33MFYGE7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7/action/storage_attestation","attest_author":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7/action/author_attestation","sign_citation":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7/action/citation_signature","submit_replication":"https://pith.science/pith/GNAYXLFD3NAFAYUBOM33MFYGE7/action/replication_record"}},"created_at":"2026-06-23T02:13:44.569239+00:00","updated_at":"2026-06-23T02:13:44.569239+00:00"}