{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:B2SJKFVPAZS3KF72TNHMEB6TUQ","short_pith_number":"pith:B2SJKFVP","schema_version":"1.0","canonical_sha256":"0ea49516af0665b517fa9b4ec207d3a413ffbf0ac6c24be18e9b328997ed0a2a","source":{"kind":"arxiv","id":"2503.06040","version":1},"attestation_state":"computed","paper":{"title":"Mitigating Memorization in LLMs using Activation Steering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amisha Bhaskar, Manan Suri, Nishit Anand","submitted_at":"2025-03-08T03:37:07Z","abstract_excerpt":"The memorization of training data by Large Language Models (LLMs) poses significant risks, including privacy leaks and the regurgitation of copyrighted content. Activation steering, a technique that directly intervenes in model activations, has emerged as a promising approach for manipulating LLMs. In this work, we explore the effectiveness of activation steering in reducing memorization while preserving generalization capabilities. We conduct empirical evaluations using a controlled memorization benchmark of literary material and demonstrate that our method successfully suppresses memorized c"},"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":"2503.06040","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-08T03:37:07Z","cross_cats_sorted":[],"title_canon_sha256":"089a130a2255e107de6f94e194b88b898291e7eae8587ad7f3cfc5e802c0a150","abstract_canon_sha256":"958afca50f821b8caf64a406c817e73c8781579052e0eff36677bf1007ba5b1d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:26:52.530736Z","signature_b64":"V7kPy+3XANj8+nuvPsLc8HFzCa/B80Cg4kZ0a+I8hZ/61Rv3ljVH345qM6vnaQYSBpFbwOd+Ryxlc0A+t7OSBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0ea49516af0665b517fa9b4ec207d3a413ffbf0ac6c24be18e9b328997ed0a2a","last_reissued_at":"2026-07-05T10:26:52.530212Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:26:52.530212Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating Memorization in LLMs using Activation Steering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amisha Bhaskar, Manan Suri, Nishit Anand","submitted_at":"2025-03-08T03:37:07Z","abstract_excerpt":"The memorization of training data by Large Language Models (LLMs) poses significant risks, including privacy leaks and the regurgitation of copyrighted content. Activation steering, a technique that directly intervenes in model activations, has emerged as a promising approach for manipulating LLMs. In this work, we explore the effectiveness of activation steering in reducing memorization while preserving generalization capabilities. We conduct empirical evaluations using a controlled memorization benchmark of literary material and demonstrate that our method successfully suppresses memorized c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.06040","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/2503.06040/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":"2503.06040","created_at":"2026-07-05T10:26:52.530270+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.06040v1","created_at":"2026-07-05T10:26:52.530270+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.06040","created_at":"2026-07-05T10:26:52.530270+00:00"},{"alias_kind":"pith_short_12","alias_value":"B2SJKFVPAZS3","created_at":"2026-07-05T10:26:52.530270+00:00"},{"alias_kind":"pith_short_16","alias_value":"B2SJKFVPAZS3KF72","created_at":"2026-07-05T10:26:52.530270+00:00"},{"alias_kind":"pith_short_8","alias_value":"B2SJKFVP","created_at":"2026-07-05T10:26:52.530270+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18767","citing_title":"Output Vector Editing for Memorization Mitigation in Large Language Models","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ","json":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ.json","graph_json":"https://pith.science/api/pith-number/B2SJKFVPAZS3KF72TNHMEB6TUQ/graph.json","events_json":"https://pith.science/api/pith-number/B2SJKFVPAZS3KF72TNHMEB6TUQ/events.json","paper":"https://pith.science/paper/B2SJKFVP"},"agent_actions":{"view_html":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ","download_json":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ.json","view_paper":"https://pith.science/paper/B2SJKFVP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.06040&json=true","fetch_graph":"https://pith.science/api/pith-number/B2SJKFVPAZS3KF72TNHMEB6TUQ/graph.json","fetch_events":"https://pith.science/api/pith-number/B2SJKFVPAZS3KF72TNHMEB6TUQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ/action/storage_attestation","attest_author":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ/action/author_attestation","sign_citation":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ/action/citation_signature","submit_replication":"https://pith.science/pith/B2SJKFVPAZS3KF72TNHMEB6TUQ/action/replication_record"}},"created_at":"2026-07-05T10:26:52.530270+00:00","updated_at":"2026-07-05T10:26:52.530270+00:00"}