{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:MHLRICJCYBWGUMBYU26PIXO4JI","short_pith_number":"pith:MHLRICJC","schema_version":"1.0","canonical_sha256":"61d7140922c06c6a3038a6bcf45ddc4a3eb18917f0125c1d7f0ec9b6da4754cf","source":{"kind":"arxiv","id":"2606.18627","version":1},"attestation_state":"computed","paper":{"title":"PACT: Preserving Anchored Cores in Task-vectors for Model Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chunyan Miao, Hao Wang, Ningyuan Shi, Peilin Zhao, Zhipeng Zhou","submitted_at":"2026-06-17T02:48:35Z","abstract_excerpt":"Model merging has emerged as a training-free alternative to multi-task learning, aiming to combine multiple task-specific fine-tuned models into a single multi-task model. Most existing model merging approaches follow the Task Arithmetic paradigm, which decomposes fine-tuned weights into pre-trained parameters and task vectors, and performs merging exclusively in the task-vector space. The effectiveness of this paradigm implicitly relies on the assumption that task-specific knowledge is encoded solely within task vectors. We argue that this assumption generally does not hold due to the intrins"},"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":"2606.18627","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-17T02:48:35Z","cross_cats_sorted":[],"title_canon_sha256":"062d32ce05b52070fb0fe8e802e242f0c692ea07a0b0006828903e937cc69499","abstract_canon_sha256":"7f87da7eae7c313ec2ea6b70f1ec3520068d157d7f838ad2fb8916d7b0e7b957"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:11:42.870470Z","signature_b64":"AOTLyrpL237zX5skS4GRQGkz3+UckDu+68FWEfLjDpjOmGnenVpfQUMJ7KmVyJitLNx474fPo7Fs7TCfqQ1pAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"61d7140922c06c6a3038a6bcf45ddc4a3eb18917f0125c1d7f0ec9b6da4754cf","last_reissued_at":"2026-06-19T16:11:42.870051Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:11:42.870051Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PACT: Preserving Anchored Cores in Task-vectors for Model Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chunyan Miao, Hao Wang, Ningyuan Shi, Peilin Zhao, Zhipeng Zhou","submitted_at":"2026-06-17T02:48:35Z","abstract_excerpt":"Model merging has emerged as a training-free alternative to multi-task learning, aiming to combine multiple task-specific fine-tuned models into a single multi-task model. Most existing model merging approaches follow the Task Arithmetic paradigm, which decomposes fine-tuned weights into pre-trained parameters and task vectors, and performs merging exclusively in the task-vector space. The effectiveness of this paradigm implicitly relies on the assumption that task-specific knowledge is encoded solely within task vectors. We argue that this assumption generally does not hold due to the intrins"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.18627","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/2606.18627/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":"2606.18627","created_at":"2026-06-19T16:11:42.870114+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.18627v1","created_at":"2026-06-19T16:11:42.870114+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.18627","created_at":"2026-06-19T16:11:42.870114+00:00"},{"alias_kind":"pith_short_12","alias_value":"MHLRICJCYBWG","created_at":"2026-06-19T16:11:42.870114+00:00"},{"alias_kind":"pith_short_16","alias_value":"MHLRICJCYBWGUMBY","created_at":"2026-06-19T16:11:42.870114+00:00"},{"alias_kind":"pith_short_8","alias_value":"MHLRICJC","created_at":"2026-06-19T16:11:42.870114+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI","json":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI.json","graph_json":"https://pith.science/api/pith-number/MHLRICJCYBWGUMBYU26PIXO4JI/graph.json","events_json":"https://pith.science/api/pith-number/MHLRICJCYBWGUMBYU26PIXO4JI/events.json","paper":"https://pith.science/paper/MHLRICJC"},"agent_actions":{"view_html":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI","download_json":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI.json","view_paper":"https://pith.science/paper/MHLRICJC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.18627&json=true","fetch_graph":"https://pith.science/api/pith-number/MHLRICJCYBWGUMBYU26PIXO4JI/graph.json","fetch_events":"https://pith.science/api/pith-number/MHLRICJCYBWGUMBYU26PIXO4JI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI/action/storage_attestation","attest_author":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI/action/author_attestation","sign_citation":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI/action/citation_signature","submit_replication":"https://pith.science/pith/MHLRICJCYBWGUMBYU26PIXO4JI/action/replication_record"}},"created_at":"2026-06-19T16:11:42.870114+00:00","updated_at":"2026-06-19T16:11:42.870114+00:00"}