{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FAVTA3T4XFBLI4TWKVCQISDRTI","short_pith_number":"pith:FAVTA3T4","schema_version":"1.0","canonical_sha256":"282b306e7cb942b4727655450448719a1097244e70dca400f064a2173fad4a3c","source":{"kind":"arxiv","id":"2410.14389","version":1},"attestation_state":"computed","paper":{"title":"SurgeryV2: Bridging the Gap Between Model Merging and Multi-Task Learning with Deep Representation Surgery","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Dacheng Tao, Enneng Yang, Guibing Guo, Jie Zhang, Li Shen, Xiaocun Cao, Xingwei Wang, Zhenyi Wang","submitted_at":"2024-10-18T11:49:40Z","abstract_excerpt":"Model merging-based multitask learning (MTL) offers a promising approach for performing MTL by merging multiple expert models without requiring access to raw training data. However, in this paper, we examine the merged model's representation distribution and uncover a critical issue of \"representation bias\". This bias arises from a significant distribution gap between the representations of the merged and expert models, leading to the suboptimal performance of the merged MTL model. To address this challenge, we first propose a representation surgery solution called Surgery. Surgery is a lightw"},"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":"2410.14389","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-18T11:49:40Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"726d16be5ba5472a9ad0df70f34e7ed5ee5e357bcd9029657ca16714be87bf98","abstract_canon_sha256":"0ae4e1d451cfa538a5963b69fb5879b1df6955334e5b6c275d91e1f0cea17fe2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:22:31.788578Z","signature_b64":"uK6nJBtua0QKSB2PXtnBYLq0MME8ZcNFuQhfenyh7P5ve3D9qCwZrBpvpPWa+CVCQJKHq4MqfS9uonbSovUxCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"282b306e7cb942b4727655450448719a1097244e70dca400f064a2173fad4a3c","last_reissued_at":"2026-07-05T09:22:31.788081Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:22:31.788081Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SurgeryV2: Bridging the Gap Between Model Merging and Multi-Task Learning with Deep Representation Surgery","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Dacheng Tao, Enneng Yang, Guibing Guo, Jie Zhang, Li Shen, Xiaocun Cao, Xingwei Wang, Zhenyi Wang","submitted_at":"2024-10-18T11:49:40Z","abstract_excerpt":"Model merging-based multitask learning (MTL) offers a promising approach for performing MTL by merging multiple expert models without requiring access to raw training data. However, in this paper, we examine the merged model's representation distribution and uncover a critical issue of \"representation bias\". This bias arises from a significant distribution gap between the representations of the merged and expert models, leading to the suboptimal performance of the merged MTL model. To address this challenge, we first propose a representation surgery solution called Surgery. Surgery is a lightw"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.14389","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/2410.14389/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":"2410.14389","created_at":"2026-07-05T09:22:31.788136+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.14389v1","created_at":"2026-07-05T09:22:31.788136+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.14389","created_at":"2026-07-05T09:22:31.788136+00:00"},{"alias_kind":"pith_short_12","alias_value":"FAVTA3T4XFBL","created_at":"2026-07-05T09:22:31.788136+00:00"},{"alias_kind":"pith_short_16","alias_value":"FAVTA3T4XFBLI4TW","created_at":"2026-07-05T09:22:31.788136+00:00"},{"alias_kind":"pith_short_8","alias_value":"FAVTA3T4","created_at":"2026-07-05T09:22:31.788136+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2408.07666","citing_title":"Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunities","ref_index":270,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13030","citing_title":"FeatCal: Feature Calibration for Post-Merging Models","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI","json":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI.json","graph_json":"https://pith.science/api/pith-number/FAVTA3T4XFBLI4TWKVCQISDRTI/graph.json","events_json":"https://pith.science/api/pith-number/FAVTA3T4XFBLI4TWKVCQISDRTI/events.json","paper":"https://pith.science/paper/FAVTA3T4"},"agent_actions":{"view_html":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI","download_json":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI.json","view_paper":"https://pith.science/paper/FAVTA3T4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.14389&json=true","fetch_graph":"https://pith.science/api/pith-number/FAVTA3T4XFBLI4TWKVCQISDRTI/graph.json","fetch_events":"https://pith.science/api/pith-number/FAVTA3T4XFBLI4TWKVCQISDRTI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI/action/storage_attestation","attest_author":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI/action/author_attestation","sign_citation":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI/action/citation_signature","submit_replication":"https://pith.science/pith/FAVTA3T4XFBLI4TWKVCQISDRTI/action/replication_record"}},"created_at":"2026-07-05T09:22:31.788136+00:00","updated_at":"2026-07-05T09:22:31.788136+00:00"}