{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:22KH7ACB2A5S3F7FX374ZSN4W4","short_pith_number":"pith:22KH7ACB","schema_version":"1.0","canonical_sha256":"d6947f8041d03b2d97e5beffccc9bcb700a28fd28bd3096733a3963ca862cfba","source":{"kind":"arxiv","id":"2112.14397","version":2},"attestation_state":"computed","paper":{"title":"EvoMoE: An Evolutional Mixture-of-Experts Training Framework via Dense-To-Sparse Gate","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bin Cui, Jilong Xue, Lingxiao Ma, Qibin Liu, Shijie Cao, Xiaonan Nie, Xupeng Miao, Yi Liu, Youshan Miao, Zhi Yang","submitted_at":"2021-12-29T04:52:14Z","abstract_excerpt":"Mixture-of-experts (MoE) is becoming popular due to its success in improving the model quality, especially in Transformers. By routing tokens with a sparse gate to a few experts (i.e., a small pieces of the full model), MoE can easily increase the model parameters to a very large scale while keeping the computation cost in a constant level. Most existing works just initialize some random experts, set a fixed gating strategy (e.g., Top-k), and train the model from scratch in an ad-hoc way. We identify that these MoE models are suffering from the immature experts and unstable sparse gate, which "},"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":"2112.14397","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-12-29T04:52:14Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"52342fb3fc26146aa896f64fc5a6d8939059391652c4aae8f93d6de75cfad74f","abstract_canon_sha256":"e8dff3835e9daea0ea1a3e6a9281a0155952b8c63fdfb693ed1c817c89f5b619"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:04:26.737671Z","signature_b64":"TJYTkff/5/7ThH31dTKFXawso88OJmk3WlNDIWcr2xPveeXH4RmlT6/8w/ADTc8375DF5Kwt8Q830ekESU2WBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d6947f8041d03b2d97e5beffccc9bcb700a28fd28bd3096733a3963ca862cfba","last_reissued_at":"2026-07-05T05:04:26.737156Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:04:26.737156Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EvoMoE: An Evolutional Mixture-of-Experts Training Framework via Dense-To-Sparse Gate","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bin Cui, Jilong Xue, Lingxiao Ma, Qibin Liu, Shijie Cao, Xiaonan Nie, Xupeng Miao, Yi Liu, Youshan Miao, Zhi Yang","submitted_at":"2021-12-29T04:52:14Z","abstract_excerpt":"Mixture-of-experts (MoE) is becoming popular due to its success in improving the model quality, especially in Transformers. By routing tokens with a sparse gate to a few experts (i.e., a small pieces of the full model), MoE can easily increase the model parameters to a very large scale while keeping the computation cost in a constant level. Most existing works just initialize some random experts, set a fixed gating strategy (e.g., Top-k), and train the model from scratch in an ad-hoc way. We identify that these MoE models are suffering from the immature experts and unstable sparse gate, which "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.14397","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/2112.14397/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":"2112.14397","created_at":"2026-07-05T05:04:26.737216+00:00"},{"alias_kind":"arxiv_version","alias_value":"2112.14397v2","created_at":"2026-07-05T05:04:26.737216+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.14397","created_at":"2026-07-05T05:04:26.737216+00:00"},{"alias_kind":"pith_short_12","alias_value":"22KH7ACB2A5S","created_at":"2026-07-05T05:04:26.737216+00:00"},{"alias_kind":"pith_short_16","alias_value":"22KH7ACB2A5S3F7F","created_at":"2026-07-05T05:04:26.737216+00:00"},{"alias_kind":"pith_short_8","alias_value":"22KH7ACB","created_at":"2026-07-05T05:04:26.737216+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20544","citing_title":"Toward Calibrated Mixture-of-Experts Under Distribution Shift","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26776","citing_title":"Towards Generalization-Oriented Models for Vehicle Routing Problems with Mixture-of-Experts","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15484","citing_title":"When Does Sparse MoE Help in Vision? The Role of Backbone Compute Leverage in Sparse Routing","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4","json":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4.json","graph_json":"https://pith.science/api/pith-number/22KH7ACB2A5S3F7FX374ZSN4W4/graph.json","events_json":"https://pith.science/api/pith-number/22KH7ACB2A5S3F7FX374ZSN4W4/events.json","paper":"https://pith.science/paper/22KH7ACB"},"agent_actions":{"view_html":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4","download_json":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4.json","view_paper":"https://pith.science/paper/22KH7ACB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2112.14397&json=true","fetch_graph":"https://pith.science/api/pith-number/22KH7ACB2A5S3F7FX374ZSN4W4/graph.json","fetch_events":"https://pith.science/api/pith-number/22KH7ACB2A5S3F7FX374ZSN4W4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4/action/storage_attestation","attest_author":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4/action/author_attestation","sign_citation":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4/action/citation_signature","submit_replication":"https://pith.science/pith/22KH7ACB2A5S3F7FX374ZSN4W4/action/replication_record"}},"created_at":"2026-07-05T05:04:26.737216+00:00","updated_at":"2026-07-05T05:04:26.737216+00:00"}