{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KLWK4FMAXF4ASRPJJGWTAEF3RR","short_pith_number":"pith:KLWK4FMA","schema_version":"1.0","canonical_sha256":"52ecae1580b9780945e949ad3010bb8c66837db61425fda4ad66a2ed8bdbf389","source":{"kind":"arxiv","id":"2412.00081","version":3},"attestation_state":"computed","paper":{"title":"Task Singular Vectors: Reducing Task Interference in Model Merging","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Antonio Andrea Gargiulo, Donato Crisostomi, Emanuele Rodol\\`a, Fabrizio Silvestri, Maria Sofia Bucarelli, Simone Scardapane","submitted_at":"2024-11-26T22:53:06Z","abstract_excerpt":"Task Arithmetic has emerged as a simple yet effective method to merge models without additional training. However, by treating entire networks as flat parameter vectors, it overlooks key structural information and is susceptible to task interference. In this paper, we study task vectors at the layer level, focusing on task layer matrices and their singular value decomposition. In particular, we concentrate on the resulting singular vectors, which we refer to as Task Singular Vectors (TSV). Recognizing that layer task matrices are often low-rank, we propose TSV-Compress (TSV-C), a simple proced"},"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":"2412.00081","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-26T22:53:06Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"c61a6a9a67b45fd106f8f6890d9e732592986908befaa1813429b3e34adbaf9f","abstract_canon_sha256":"309130d859df04a7cd25502fef916dcc302276700149faa5d23a0a5ebf6d2385"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:44:20.308408Z","signature_b64":"4BOBuyrUgYmAO+4mZzxe5T6PLL1oO1mj9PqOaTCPYmdhrWAuiA3bRnj2yNEgRTHl9tiQtxIMkTgNDcYs1zi0Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"52ecae1580b9780945e949ad3010bb8c66837db61425fda4ad66a2ed8bdbf389","last_reissued_at":"2026-07-05T10:44:20.307928Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:44:20.307928Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Task Singular Vectors: Reducing Task Interference in Model Merging","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Antonio Andrea Gargiulo, Donato Crisostomi, Emanuele Rodol\\`a, Fabrizio Silvestri, Maria Sofia Bucarelli, Simone Scardapane","submitted_at":"2024-11-26T22:53:06Z","abstract_excerpt":"Task Arithmetic has emerged as a simple yet effective method to merge models without additional training. However, by treating entire networks as flat parameter vectors, it overlooks key structural information and is susceptible to task interference. In this paper, we study task vectors at the layer level, focusing on task layer matrices and their singular value decomposition. In particular, we concentrate on the resulting singular vectors, which we refer to as Task Singular Vectors (TSV). Recognizing that layer task matrices are often low-rank, we propose TSV-Compress (TSV-C), a simple proced"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.00081","kind":"arxiv","version":3},"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/2412.00081/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":"2412.00081","created_at":"2026-07-05T10:44:20.307984+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.00081v3","created_at":"2026-07-05T10:44:20.307984+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.00081","created_at":"2026-07-05T10:44:20.307984+00:00"},{"alias_kind":"pith_short_12","alias_value":"KLWK4FMAXF4A","created_at":"2026-07-05T10:44:20.307984+00:00"},{"alias_kind":"pith_short_16","alias_value":"KLWK4FMAXF4ASRPJ","created_at":"2026-07-05T10:44:20.307984+00:00"},{"alias_kind":"pith_short_8","alias_value":"KLWK4FMA","created_at":"2026-07-05T10:44:20.307984+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22589","citing_title":"Training-free Task Classification for Multi-Task Model Merging","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02252","citing_title":"ResMerge: Residual-based Spectral Merging of Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02881","citing_title":"One Model to Translate Them All? A Journey to Mount Doom for Multilingual Model Merging","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR","json":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR.json","graph_json":"https://pith.science/api/pith-number/KLWK4FMAXF4ASRPJJGWTAEF3RR/graph.json","events_json":"https://pith.science/api/pith-number/KLWK4FMAXF4ASRPJJGWTAEF3RR/events.json","paper":"https://pith.science/paper/KLWK4FMA"},"agent_actions":{"view_html":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR","download_json":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR.json","view_paper":"https://pith.science/paper/KLWK4FMA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.00081&json=true","fetch_graph":"https://pith.science/api/pith-number/KLWK4FMAXF4ASRPJJGWTAEF3RR/graph.json","fetch_events":"https://pith.science/api/pith-number/KLWK4FMAXF4ASRPJJGWTAEF3RR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR/action/storage_attestation","attest_author":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR/action/author_attestation","sign_citation":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR/action/citation_signature","submit_replication":"https://pith.science/pith/KLWK4FMAXF4ASRPJJGWTAEF3RR/action/replication_record"}},"created_at":"2026-07-05T10:44:20.307984+00:00","updated_at":"2026-07-05T10:44:20.307984+00:00"}