{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:42BL6OBC6OVWOBRP7OVO5XPPLF","short_pith_number":"pith:42BL6OBC","schema_version":"1.0","canonical_sha256":"e682bf3822f3ab67062ffbaaeeddef597bde3c4c417ba2b3aaa70fca18e1890d","source":{"kind":"arxiv","id":"2507.04250","version":1},"attestation_state":"computed","paper":{"title":"Just Enough Shifts: Mitigating Over-Refusal in Aligned Language Models with Targeted Representation Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Charles Fleming, Mahavir Dabas, Ming Jin, Ruoxi Jia, Si Chen","submitted_at":"2025-07-06T05:47:04Z","abstract_excerpt":"Safety alignment is crucial for large language models (LLMs) to resist malicious instructions but often results in over-refusals, where benign prompts are unnecessarily rejected, impairing user experience and model utility. We introduce ACTOR (Activation-Based Training for Over-Refusal Reduction), a robust and compute- and data-efficient training framework that minimizes over-refusals by leveraging internal activation patterns from diverse queries. ACTOR precisely identifies and adjusts the activation components that trigger refusals, providing stronger control over the refusal mechanism. By f"},"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":"2507.04250","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-07-06T05:47:04Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"44d73b59fe035a66f22095af86f20c1c7b98c9606bd4b923c264a29d2e026cab","abstract_canon_sha256":"f79008a3ba790a59ac19c9a897bc9e1dffeac194fec34474692e96afe961c0a3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:32:44.226900Z","signature_b64":"HBIM4YiW3WGW6XZhEI6NJ3FoPsB532vT47ZEfWNM1OCbmf39zAvxXDatjVuEGUD3iXhmwSgwrm0vEKcRoB+eCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e682bf3822f3ab67062ffbaaeeddef597bde3c4c417ba2b3aaa70fca18e1890d","last_reissued_at":"2026-07-05T11:32:44.226428Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:32:44.226428Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Just Enough Shifts: Mitigating Over-Refusal in Aligned Language Models with Targeted Representation Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Charles Fleming, Mahavir Dabas, Ming Jin, Ruoxi Jia, Si Chen","submitted_at":"2025-07-06T05:47:04Z","abstract_excerpt":"Safety alignment is crucial for large language models (LLMs) to resist malicious instructions but often results in over-refusals, where benign prompts are unnecessarily rejected, impairing user experience and model utility. We introduce ACTOR (Activation-Based Training for Over-Refusal Reduction), a robust and compute- and data-efficient training framework that minimizes over-refusals by leveraging internal activation patterns from diverse queries. ACTOR precisely identifies and adjusts the activation components that trigger refusals, providing stronger control over the refusal mechanism. By f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.04250","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/2507.04250/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":"2507.04250","created_at":"2026-07-05T11:32:44.226484+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.04250v1","created_at":"2026-07-05T11:32:44.226484+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.04250","created_at":"2026-07-05T11:32:44.226484+00:00"},{"alias_kind":"pith_short_12","alias_value":"42BL6OBC6OVW","created_at":"2026-07-05T11:32:44.226484+00:00"},{"alias_kind":"pith_short_16","alias_value":"42BL6OBC6OVWOBRP","created_at":"2026-07-05T11:32:44.226484+00:00"},{"alias_kind":"pith_short_8","alias_value":"42BL6OBC","created_at":"2026-07-05T11:32:44.226484+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21147","citing_title":"AOR-Bench: Do Large Audio Language Models Over-Refuse Pseudo-Harmful Queries?","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07335","citing_title":"Defending Jailbreak Attacks on Large Language Models via Manifold Trajectory Kinetics","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24154","citing_title":"Palette: A Modular, Controllable, and Efficient Framework for On-demand Authorized Safety Alignment Relaxation in LLMs","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11679","citing_title":"Explaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional Expansion","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11679","citing_title":"Explaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional Expansion","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17614","citing_title":"Characterizing Model-Native Skills","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF","json":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF.json","graph_json":"https://pith.science/api/pith-number/42BL6OBC6OVWOBRP7OVO5XPPLF/graph.json","events_json":"https://pith.science/api/pith-number/42BL6OBC6OVWOBRP7OVO5XPPLF/events.json","paper":"https://pith.science/paper/42BL6OBC"},"agent_actions":{"view_html":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF","download_json":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF.json","view_paper":"https://pith.science/paper/42BL6OBC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.04250&json=true","fetch_graph":"https://pith.science/api/pith-number/42BL6OBC6OVWOBRP7OVO5XPPLF/graph.json","fetch_events":"https://pith.science/api/pith-number/42BL6OBC6OVWOBRP7OVO5XPPLF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF/action/storage_attestation","attest_author":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF/action/author_attestation","sign_citation":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF/action/citation_signature","submit_replication":"https://pith.science/pith/42BL6OBC6OVWOBRP7OVO5XPPLF/action/replication_record"}},"created_at":"2026-07-05T11:32:44.226484+00:00","updated_at":"2026-07-05T11:32:44.226484+00:00"}