{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:OF56Y7Z2KKNQ47AD5FHCWKGA54","short_pith_number":"pith:OF56Y7Z2","schema_version":"1.0","canonical_sha256":"717bec7f3a529b0e7c03e94e2b28c0ef09b8e1f2d9d4730d4e59107c1553779d","source":{"kind":"arxiv","id":"2407.07089","version":2},"attestation_state":"computed","paper":{"title":"Fine-Tuning Attention Modules Only: Enhancing Weight Disentanglement in Task Arithmetic","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bojian Hou, Jiancong Xiao, Li Shen, Ruochen Jin, Weijie Su","submitted_at":"2024-07-09T17:59:17Z","abstract_excerpt":"In recent years, task arithmetic has garnered increasing attention. This approach edits pre-trained models directly in weight space by combining the fine-tuned weights of various tasks into a unified model. Its efficiency and cost-effectiveness stem from its training-free combination, contrasting with traditional methods that require model training on large datasets for multiple tasks. However, applying such a unified model to individual tasks can lead to interference from other tasks (lack of weight disentanglement). To address this issue, Neural Tangent Kernel (NTK) linearization has been em"},"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":"2407.07089","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-09T17:59:17Z","cross_cats_sorted":[],"title_canon_sha256":"9bde10182106b5ed12df781c3ca9e500c51461fdcae67528e6ad8c07d2484299","abstract_canon_sha256":"7f7af5e55b2f44bc3ed0f08260f698aafde399920a97a04f5bf994a9d2fafad5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:06:42.063424Z","signature_b64":"MvTXTAUESu1Yp6HM48s9DJ6NIrZ/cJPAlI7Z4sqp/9eMe1HTcU7Eqqg2Ju02xj7OxIoDXpntxTTWJNGQH13qBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"717bec7f3a529b0e7c03e94e2b28c0ef09b8e1f2d9d4730d4e59107c1553779d","last_reissued_at":"2026-07-05T10:06:42.062885Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:06:42.062885Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fine-Tuning Attention Modules Only: Enhancing Weight Disentanglement in Task Arithmetic","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bojian Hou, Jiancong Xiao, Li Shen, Ruochen Jin, Weijie Su","submitted_at":"2024-07-09T17:59:17Z","abstract_excerpt":"In recent years, task arithmetic has garnered increasing attention. This approach edits pre-trained models directly in weight space by combining the fine-tuned weights of various tasks into a unified model. Its efficiency and cost-effectiveness stem from its training-free combination, contrasting with traditional methods that require model training on large datasets for multiple tasks. However, applying such a unified model to individual tasks can lead to interference from other tasks (lack of weight disentanglement). To address this issue, Neural Tangent Kernel (NTK) linearization has been em"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.07089","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/2407.07089/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":"2407.07089","created_at":"2026-07-05T10:06:42.062955+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.07089v2","created_at":"2026-07-05T10:06:42.062955+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.07089","created_at":"2026-07-05T10:06:42.062955+00:00"},{"alias_kind":"pith_short_12","alias_value":"OF56Y7Z2KKNQ","created_at":"2026-07-05T10:06:42.062955+00:00"},{"alias_kind":"pith_short_16","alias_value":"OF56Y7Z2KKNQ47AD","created_at":"2026-07-05T10:06:42.062955+00:00"},{"alias_kind":"pith_short_8","alias_value":"OF56Y7Z2","created_at":"2026-07-05T10:06:42.062955+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2506.23104","citing_title":"DC-TTA: Divide-and-Conquer Framework for Test-Time Adaptation of Interactive Segmentation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2408.07666","citing_title":"Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunities","ref_index":101,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54","json":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54.json","graph_json":"https://pith.science/api/pith-number/OF56Y7Z2KKNQ47AD5FHCWKGA54/graph.json","events_json":"https://pith.science/api/pith-number/OF56Y7Z2KKNQ47AD5FHCWKGA54/events.json","paper":"https://pith.science/paper/OF56Y7Z2"},"agent_actions":{"view_html":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54","download_json":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54.json","view_paper":"https://pith.science/paper/OF56Y7Z2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.07089&json=true","fetch_graph":"https://pith.science/api/pith-number/OF56Y7Z2KKNQ47AD5FHCWKGA54/graph.json","fetch_events":"https://pith.science/api/pith-number/OF56Y7Z2KKNQ47AD5FHCWKGA54/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54/action/storage_attestation","attest_author":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54/action/author_attestation","sign_citation":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54/action/citation_signature","submit_replication":"https://pith.science/pith/OF56Y7Z2KKNQ47AD5FHCWKGA54/action/replication_record"}},"created_at":"2026-07-05T10:06:42.062955+00:00","updated_at":"2026-07-05T10:06:42.062955+00:00"}