{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4MHGIJMTID7TOARLDN6BUSHLTR","short_pith_number":"pith:4MHGIJMT","schema_version":"1.0","canonical_sha256":"e30e64259340ff37022b1b7c1a48eb9c5e427f294a58e208916cef0893690851","source":{"kind":"arxiv","id":"2406.01563","version":2},"attestation_state":"computed","paper":{"title":"LoFiT: Localized Fine-tuning on LLM Representations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fangcong Yin, Greg Durrett, Xi Ye","submitted_at":"2024-06-03T17:45:41Z","abstract_excerpt":"Recent work in interpretability shows that large language models (LLMs) can be adapted for new tasks in a learning-free way: it is possible to intervene on LLM representations to elicit desired behaviors for alignment. For instance, adding certain bias vectors to the outputs of certain attention heads is reported to boost the truthfulness of models. In this work, we show that localized fine-tuning serves as an effective alternative to such representation intervention methods. We introduce a framework called Localized Fine-Tuning on LLM Representations (LoFiT), which identifies a subset of atte"},"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":"2406.01563","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-03T17:45:41Z","cross_cats_sorted":[],"title_canon_sha256":"a125d5ab31ae15c0314eb74271a674aa535b103d22d13281803b718abfe729d7","abstract_canon_sha256":"b53f3feeb435f248fc6535e0d50b060b8b3ea633bd7117ecc7e22b701500fa95"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:50.806353Z","signature_b64":"MyKT4SppbF+CMzOsvwhdxngfx4j5PHaEWYBh8aa/eGmw+PZUR+eFxildb/TQ8yoqWQOBycYXqGkVMaoRwBJYBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e30e64259340ff37022b1b7c1a48eb9c5e427f294a58e208916cef0893690851","last_reissued_at":"2026-07-05T09:28:50.805875Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:50.805875Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoFiT: Localized Fine-tuning on LLM Representations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fangcong Yin, Greg Durrett, Xi Ye","submitted_at":"2024-06-03T17:45:41Z","abstract_excerpt":"Recent work in interpretability shows that large language models (LLMs) can be adapted for new tasks in a learning-free way: it is possible to intervene on LLM representations to elicit desired behaviors for alignment. For instance, adding certain bias vectors to the outputs of certain attention heads is reported to boost the truthfulness of models. In this work, we show that localized fine-tuning serves as an effective alternative to such representation intervention methods. We introduce a framework called Localized Fine-Tuning on LLM Representations (LoFiT), which identifies a subset of atte"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.01563","kind":"arxiv","version":2},"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/2406.01563/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":"2406.01563","created_at":"2026-07-05T09:28:50.805928+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.01563v2","created_at":"2026-07-05T09:28:50.805928+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.01563","created_at":"2026-07-05T09:28:50.805928+00:00"},{"alias_kind":"pith_short_12","alias_value":"4MHGIJMTID7T","created_at":"2026-07-05T09:28:50.805928+00:00"},{"alias_kind":"pith_short_16","alias_value":"4MHGIJMTID7TOARL","created_at":"2026-07-05T09:28:50.805928+00:00"},{"alias_kind":"pith_short_8","alias_value":"4MHGIJMT","created_at":"2026-07-05T09:28:50.805928+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07812","citing_title":"Scaling Participation in Modular AI Systems","ref_index":141,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14174","citing_title":"Correcting Suppressed Log-Probabilities in Language Models with Post-Transformer Adapters","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06393","citing_title":"ART: Attention Replacement Technique to Improve Factuality in LLMs","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR","json":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR.json","graph_json":"https://pith.science/api/pith-number/4MHGIJMTID7TOARLDN6BUSHLTR/graph.json","events_json":"https://pith.science/api/pith-number/4MHGIJMTID7TOARLDN6BUSHLTR/events.json","paper":"https://pith.science/paper/4MHGIJMT"},"agent_actions":{"view_html":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR","download_json":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR.json","view_paper":"https://pith.science/paper/4MHGIJMT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.01563&json=true","fetch_graph":"https://pith.science/api/pith-number/4MHGIJMTID7TOARLDN6BUSHLTR/graph.json","fetch_events":"https://pith.science/api/pith-number/4MHGIJMTID7TOARLDN6BUSHLTR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR/action/storage_attestation","attest_author":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR/action/author_attestation","sign_citation":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR/action/citation_signature","submit_replication":"https://pith.science/pith/4MHGIJMTID7TOARLDN6BUSHLTR/action/replication_record"}},"created_at":"2026-07-05T09:28:50.805928+00:00","updated_at":"2026-07-05T09:28:50.805928+00:00"}