{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5CNFGOSXPBX3N7VV5IKMNNUK6E","short_pith_number":"pith:5CNFGOSX","schema_version":"1.0","canonical_sha256":"e89a533a57786fb6feb5ea14c6b68af10e90c81e6841013c97ba7290f6e72d9c","source":{"kind":"arxiv","id":"2603.06626","version":2},"attestation_state":"computed","paper":{"title":"Grouter: Decoupling Routing from Representation for Accelerated MoE Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Kun Yuan, Mou Sun, Rizhen Hu, Yuqi Xu, Zihan Liu","submitted_at":"2026-02-22T06:09:57Z","abstract_excerpt":"Traditional Mixture-of-Experts (MoE) training typically proceeds without any structural priors, effectively requiring the model to simultaneously train expert weights while searching for an optimal routing policy within a vast combinatorial space. This entanglement often leads to sluggish convergence and training instabilities. This paper introduces Grouter, a preemptive routing method that by distilling high-quality structures from fully-trained MoE models and serving as a fixed router for target models. By decoupling structural optimization from weight updates, Grouter significantly accelera"},"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":"2603.06626","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-22T06:09:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"09153dee396a77785b729570f6cf58d9f16293d5e551ec9e4be0a08790ce60a1","abstract_canon_sha256":"3ee0b9716a3087c4bc01000847ed67fba7af11b7aee42272b649ad2b7344e425"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:27.233962Z","signature_b64":"3JmXHO3492XHKQvzCoR/PkKhd85q/TQRdaUmB0vIILjhekiCCX8eZGbApOVo7bi7UPIXN09L71parQUKZCr2Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e89a533a57786fb6feb5ea14c6b68af10e90c81e6841013c97ba7290f6e72d9c","last_reissued_at":"2026-05-26T01:03:27.233028Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:27.233028Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Grouter: Decoupling Routing from Representation for Accelerated MoE Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Kun Yuan, Mou Sun, Rizhen Hu, Yuqi Xu, Zihan Liu","submitted_at":"2026-02-22T06:09:57Z","abstract_excerpt":"Traditional Mixture-of-Experts (MoE) training typically proceeds without any structural priors, effectively requiring the model to simultaneously train expert weights while searching for an optimal routing policy within a vast combinatorial space. This entanglement often leads to sluggish convergence and training instabilities. This paper introduces Grouter, a preemptive routing method that by distilling high-quality structures from fully-trained MoE models and serving as a fixed router for target models. By decoupling structural optimization from weight updates, Grouter significantly accelera"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.06626","kind":"arxiv","version":2},"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/2603.06626/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":"2603.06626","created_at":"2026-05-26T01:03:27.233161+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.06626v2","created_at":"2026-05-26T01:03:27.233161+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.06626","created_at":"2026-05-26T01:03:27.233161+00:00"},{"alias_kind":"pith_short_12","alias_value":"5CNFGOSXPBX3","created_at":"2026-05-26T01:03:27.233161+00:00"},{"alias_kind":"pith_short_16","alias_value":"5CNFGOSXPBX3N7VV","created_at":"2026-05-26T01:03:27.233161+00:00"},{"alias_kind":"pith_short_8","alias_value":"5CNFGOSX","created_at":"2026-05-26T01:03:27.233161+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2605.17333","citing_title":"Leveraging Error Diversity in Group Rollouts for Reinforcement Learning","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17291","citing_title":"Step-wise Rubric Rewards for LLM Reasoning","ref_index":22,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12476","citing_title":"Routers Learn the Geometry of Their Experts: Geometric Coupling in Sparse Mixture-of-Experts","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2604.14419","citing_title":"Equifinality in Mixture of Experts: Routing Topology Does Not Determine Language Modeling Quality","ref_index":32,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E","json":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E.json","graph_json":"https://pith.science/api/pith-number/5CNFGOSXPBX3N7VV5IKMNNUK6E/graph.json","events_json":"https://pith.science/api/pith-number/5CNFGOSXPBX3N7VV5IKMNNUK6E/events.json","paper":"https://pith.science/paper/5CNFGOSX"},"agent_actions":{"view_html":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E","download_json":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E.json","view_paper":"https://pith.science/paper/5CNFGOSX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.06626&json=true","fetch_graph":"https://pith.science/api/pith-number/5CNFGOSXPBX3N7VV5IKMNNUK6E/graph.json","fetch_events":"https://pith.science/api/pith-number/5CNFGOSXPBX3N7VV5IKMNNUK6E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E/action/storage_attestation","attest_author":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E/action/author_attestation","sign_citation":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E/action/citation_signature","submit_replication":"https://pith.science/pith/5CNFGOSXPBX3N7VV5IKMNNUK6E/action/replication_record"}},"created_at":"2026-05-26T01:03:27.233161+00:00","updated_at":"2026-05-26T01:03:27.233161+00:00"}