{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2DXJJBOCZLZRA7YNGG5JQTFFRK","short_pith_number":"pith:2DXJJBOC","schema_version":"1.0","canonical_sha256":"d0ee9485c2caf3107f0d31ba984ca58a9160ebc3faad02250b6b61aa3cbb4290","source":{"kind":"arxiv","id":"2608.08627","version":1},"attestation_state":"computed","paper":{"title":"UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bin Gu, Changjiang Jiang, Chao Huang, Fanhu Zeng, Jianbo Zhao, Lei Xin, Peize Li, Xuyang Zhao, Yanyue Xie, Zhenglun Kong, Zitong Wang, Zunhai Su","submitted_at":"2026-08-09T10:20:20Z","abstract_excerpt":"Sparse mixture-of-experts (MoE) layers expand recommendation capacity through conditional computation, yet a trained checkpoint still stores and routes over its full expert bank. We study a deployment problem: convert that checkpoint to a smaller standard MoE under an explicit expert budget, without adding a compression-specific online module. To address this, we introduce UniMoMo, a post-training compression framework formulated as a constrained graph coarsening problem. Rather than relying on parameter distance, UniMoMo groups experts based on their functional similarity, using an unlabeled "},"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":"2608.08627","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-08-09T10:20:20Z","cross_cats_sorted":[],"title_canon_sha256":"619e781ed3ab0fd96bd281e77f0e1cf2d726169e4de04af2d1d6ea266c93357a","abstract_canon_sha256":"0e03fe664bab1e2d1425ae7018a63cc34eed73d5520be32b67cbca33263936a9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T01:23:12.674162Z","signature_b64":"Hs+zcqJ7h1ghMOQlzYX+VlMBpIOfoOk+pvFaTi6TllY6c1FqrLE0c1Q7ckdOq7Qs0kaEWkLIOxhZfzAWWOQQDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d0ee9485c2caf3107f0d31ba984ca58a9160ebc3faad02250b6b61aa3cbb4290","last_reissued_at":"2026-08-11T01:23:12.671673Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T01:23:12.671673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bin Gu, Changjiang Jiang, Chao Huang, Fanhu Zeng, Jianbo Zhao, Lei Xin, Peize Li, Xuyang Zhao, Yanyue Xie, Zhenglun Kong, Zitong Wang, Zunhai Su","submitted_at":"2026-08-09T10:20:20Z","abstract_excerpt":"Sparse mixture-of-experts (MoE) layers expand recommendation capacity through conditional computation, yet a trained checkpoint still stores and routes over its full expert bank. We study a deployment problem: convert that checkpoint to a smaller standard MoE under an explicit expert budget, without adding a compression-specific online module. To address this, we introduce UniMoMo, a post-training compression framework formulated as a constrained graph coarsening problem. Rather than relying on parameter distance, UniMoMo groups experts based on their functional similarity, using an unlabeled "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.08627","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/2608.08627/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":"2608.08627","created_at":"2026-08-11T01:23:12.672486+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.08627v1","created_at":"2026-08-11T01:23:12.672486+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.08627","created_at":"2026-08-11T01:23:12.672486+00:00"},{"alias_kind":"pith_short_12","alias_value":"2DXJJBOCZLZR","created_at":"2026-08-11T01:23:12.672486+00:00"},{"alias_kind":"pith_short_16","alias_value":"2DXJJBOCZLZRA7YN","created_at":"2026-08-11T01:23:12.672486+00:00"},{"alias_kind":"pith_short_8","alias_value":"2DXJJBOC","created_at":"2026-08-11T01:23:12.672486+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/2DXJJBOCZLZRA7YNGG5JQTFFRK","json":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK.json","graph_json":"https://pith.science/api/pith-number/2DXJJBOCZLZRA7YNGG5JQTFFRK/graph.json","events_json":"https://pith.science/api/pith-number/2DXJJBOCZLZRA7YNGG5JQTFFRK/events.json","paper":"https://pith.science/paper/2DXJJBOC"},"agent_actions":{"view_html":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK","download_json":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK.json","view_paper":"https://pith.science/paper/2DXJJBOC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.08627&json=true","fetch_graph":"https://pith.science/api/pith-number/2DXJJBOCZLZRA7YNGG5JQTFFRK/graph.json","fetch_events":"https://pith.science/api/pith-number/2DXJJBOCZLZRA7YNGG5JQTFFRK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK/action/storage_attestation","attest_author":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK/action/author_attestation","sign_citation":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK/action/citation_signature","submit_replication":"https://pith.science/pith/2DXJJBOCZLZRA7YNGG5JQTFFRK/action/replication_record"}},"created_at":"2026-08-11T01:23:12.672486+00:00","updated_at":"2026-08-11T01:23:12.672486+00:00"}