{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:A2OWGNNDMRJTM6SAN7DNQRUBU7","short_pith_number":"pith:A2OWGNND","schema_version":"1.0","canonical_sha256":"069d6335a36453367a406fc6d84681a7c9cb4d09bd4eafc53077a1e7c1606b8c","source":{"kind":"arxiv","id":"2605.29280","version":1},"attestation_state":"computed","paper":{"title":"LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.LG","authors_text":"Boyang Liu, Can Cui, Chenxiao Guan, Chuanqi Xu, Darren Liu, Dharak Kharod, Ellie Dingqiao Wen, Erzhuo Wang, Gerard Jonathan Mugisha Akkerhuis, Haicheng Chen, Huayu Li, Hua Zheng, Hui Zhou, Jiyan Yang, Kenny Lov, Laming Chen, Lisang Ding, Parish Aggarwal, Ping Chen, Qiang Jin, Qiao Yang, Qinghai Zhou, Qin Huang, Rong Jin, Ruichao Qiu, Rui Yang, Santanu Kolay, Shali Jiang, Shifu Xu, Shuo Chang, Shuyu Xu, Song Zhou, Wankun Zhu, Wenlin Chen, Xian Chen, Xiangge Meng, Xiaolong Liu, Xiaorui Gan, Xiaoyi Liu, Xin Xu, Yasmine Badr, Yuzhen Huang, Zhehui Zhou","submitted_at":"2026-05-28T02:59:46Z","abstract_excerpt":"Knowledge distillation (KD) transfers a single scalar prediction from a large foundation model (FM) to compact vertical models (VMs), suffering from diminishing transfer ratio -- the fraction of FM improvement captured by the VM -- as a single scalar cannot convey the rich intermediate knowledge that larger FMs learn. To address this bottleneck, we propose LoopFM (Learning frOm HistOrical ReP*resentations of FM), a framework that opens a high-bandwidth transfer channel by structuring FM intermediate embeddings as input features (e.g., user history sequence) for downstream VMs, without requirin"},"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":"2605.29280","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-28T02:59:46Z","cross_cats_sorted":["cs.AI","cs.IR"],"title_canon_sha256":"11296fb3ec8f261dcf48878ea54dc5d08bc520bce9954771ca9d171784ea226f","abstract_canon_sha256":"098344714f7f3f519495fc58e0501d45340413a56681bc12c03c9b09e4629f8d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:29.194428Z","signature_b64":"r45KjR4iZdC/mqNYyvaKqQW3lNzNJmOQ/LPPRSJgdu3y3VEGGqH+GPuA8C5HECKk9N/pngfBXbNGtXwOUZyaCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"069d6335a36453367a406fc6d84681a7c9cb4d09bd4eafc53077a1e7c1606b8c","last_reissued_at":"2026-05-29T01:05:29.193717Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:29.193717Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.LG","authors_text":"Boyang Liu, Can Cui, Chenxiao Guan, Chuanqi Xu, Darren Liu, Dharak Kharod, Ellie Dingqiao Wen, Erzhuo Wang, Gerard Jonathan Mugisha Akkerhuis, Haicheng Chen, Huayu Li, Hua Zheng, Hui Zhou, Jiyan Yang, Kenny Lov, Laming Chen, Lisang Ding, Parish Aggarwal, Ping Chen, Qiang Jin, Qiao Yang, Qinghai Zhou, Qin Huang, Rong Jin, Ruichao Qiu, Rui Yang, Santanu Kolay, Shali Jiang, Shifu Xu, Shuo Chang, Shuyu Xu, Song Zhou, Wankun Zhu, Wenlin Chen, Xian Chen, Xiangge Meng, Xiaolong Liu, Xiaorui Gan, Xiaoyi Liu, Xin Xu, Yasmine Badr, Yuzhen Huang, Zhehui Zhou","submitted_at":"2026-05-28T02:59:46Z","abstract_excerpt":"Knowledge distillation (KD) transfers a single scalar prediction from a large foundation model (FM) to compact vertical models (VMs), suffering from diminishing transfer ratio -- the fraction of FM improvement captured by the VM -- as a single scalar cannot convey the rich intermediate knowledge that larger FMs learn. To address this bottleneck, we propose LoopFM (Learning frOm HistOrical ReP*resentations of FM), a framework that opens a high-bandwidth transfer channel by structuring FM intermediate embeddings as input features (e.g., user history sequence) for downstream VMs, without requirin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29280","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/2605.29280/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":"2605.29280","created_at":"2026-05-29T01:05:29.193801+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29280v1","created_at":"2026-05-29T01:05:29.193801+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29280","created_at":"2026-05-29T01:05:29.193801+00:00"},{"alias_kind":"pith_short_12","alias_value":"A2OWGNNDMRJT","created_at":"2026-05-29T01:05:29.193801+00:00"},{"alias_kind":"pith_short_16","alias_value":"A2OWGNNDMRJTM6SA","created_at":"2026-05-29T01:05:29.193801+00:00"},{"alias_kind":"pith_short_8","alias_value":"A2OWGNND","created_at":"2026-05-29T01:05:29.193801+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/A2OWGNNDMRJTM6SAN7DNQRUBU7","json":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7.json","graph_json":"https://pith.science/api/pith-number/A2OWGNNDMRJTM6SAN7DNQRUBU7/graph.json","events_json":"https://pith.science/api/pith-number/A2OWGNNDMRJTM6SAN7DNQRUBU7/events.json","paper":"https://pith.science/paper/A2OWGNND"},"agent_actions":{"view_html":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7","download_json":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7.json","view_paper":"https://pith.science/paper/A2OWGNND","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29280&json=true","fetch_graph":"https://pith.science/api/pith-number/A2OWGNNDMRJTM6SAN7DNQRUBU7/graph.json","fetch_events":"https://pith.science/api/pith-number/A2OWGNNDMRJTM6SAN7DNQRUBU7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7/action/storage_attestation","attest_author":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7/action/author_attestation","sign_citation":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7/action/citation_signature","submit_replication":"https://pith.science/pith/A2OWGNNDMRJTM6SAN7DNQRUBU7/action/replication_record"}},"created_at":"2026-05-29T01:05:29.193801+00:00","updated_at":"2026-05-29T01:05:29.193801+00:00"}