{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:MQ2BJARDMWEKO5GYCGPRU2INUG","short_pith_number":"pith:MQ2BJARD","schema_version":"1.0","canonical_sha256":"64341482236588a774d8119f1a690da1940ba3ef083e3b90a13a8c33f4a884cb","source":{"kind":"arxiv","id":"2607.11475","version":1},"attestation_state":"computed","paper":{"title":"HyperSafe: Inference-Time Safety Recovery for Fine-Tuned Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Abdelrahman Eldesokey, Aznaur Aliev, Bang An, Bernard Ghanem, Carlos Hinojosa, Yibo Yang","submitted_at":"2026-07-13T12:28:43Z","abstract_excerpt":"Safety alignment in large language models can be fragile under fine-tuning, as even benign task adaptation may increase harmful compliance. Existing defenses mainly follow two directions: they either intervene during or after fine-tuning through retraining or weight modification, which can be costly and may hurt task performance, or they use model-agnostic safety classifiers, which may miss failures specific to a given fine-tuned checkpoint. These limitations motivate a post hoc, model-specific, and non-invasive approach to safety restoration. To meet these requirements, we propose HyperSafe, "},"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":"2607.11475","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-07-13T12:28:43Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2674afaf73535d12fcb574dc8c72414a71b9fa1d31f8faa9aa98080b48354a89","abstract_canon_sha256":"d66d617432f4f3c7b22fe6f44564d7b0c5327ff6638efee18a3ce2168a708388"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T02:22:05.098004Z","signature_b64":"KLHeRRIcXbyn0vT84wfnbJFnngw+ucgTdQ2QVQlsXfRL2SX+uihSJKojcpZTOdBNz8QAMZldOe5AlEhS50pLBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"64341482236588a774d8119f1a690da1940ba3ef083e3b90a13a8c33f4a884cb","last_reissued_at":"2026-07-14T02:22:05.097062Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T02:22:05.097062Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HyperSafe: Inference-Time Safety Recovery for Fine-Tuned Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Abdelrahman Eldesokey, Aznaur Aliev, Bang An, Bernard Ghanem, Carlos Hinojosa, Yibo Yang","submitted_at":"2026-07-13T12:28:43Z","abstract_excerpt":"Safety alignment in large language models can be fragile under fine-tuning, as even benign task adaptation may increase harmful compliance. Existing defenses mainly follow two directions: they either intervene during or after fine-tuning through retraining or weight modification, which can be costly and may hurt task performance, or they use model-agnostic safety classifiers, which may miss failures specific to a given fine-tuned checkpoint. These limitations motivate a post hoc, model-specific, and non-invasive approach to safety restoration. To meet these requirements, we propose HyperSafe, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.11475","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/2607.11475/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":"2607.11475","created_at":"2026-07-14T02:22:05.097476+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.11475v1","created_at":"2026-07-14T02:22:05.097476+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.11475","created_at":"2026-07-14T02:22:05.097476+00:00"},{"alias_kind":"pith_short_12","alias_value":"MQ2BJARDMWEK","created_at":"2026-07-14T02:22:05.097476+00:00"},{"alias_kind":"pith_short_16","alias_value":"MQ2BJARDMWEKO5GY","created_at":"2026-07-14T02:22:05.097476+00:00"},{"alias_kind":"pith_short_8","alias_value":"MQ2BJARD","created_at":"2026-07-14T02:22:05.097476+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/MQ2BJARDMWEKO5GYCGPRU2INUG","json":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG.json","graph_json":"https://pith.science/api/pith-number/MQ2BJARDMWEKO5GYCGPRU2INUG/graph.json","events_json":"https://pith.science/api/pith-number/MQ2BJARDMWEKO5GYCGPRU2INUG/events.json","paper":"https://pith.science/paper/MQ2BJARD"},"agent_actions":{"view_html":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG","download_json":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG.json","view_paper":"https://pith.science/paper/MQ2BJARD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.11475&json=true","fetch_graph":"https://pith.science/api/pith-number/MQ2BJARDMWEKO5GYCGPRU2INUG/graph.json","fetch_events":"https://pith.science/api/pith-number/MQ2BJARDMWEKO5GYCGPRU2INUG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG/action/storage_attestation","attest_author":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG/action/author_attestation","sign_citation":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG/action/citation_signature","submit_replication":"https://pith.science/pith/MQ2BJARDMWEKO5GYCGPRU2INUG/action/replication_record"}},"created_at":"2026-07-14T02:22:05.097476+00:00","updated_at":"2026-07-14T02:22:05.097476+00:00"}