{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:JGJCN27PMNDRV74VTLO2LHHVZZ","short_pith_number":"pith:JGJCN27P","schema_version":"1.0","canonical_sha256":"499226ebef63471aff959adda59cf5ce58ab4090da69114215f70c44f2ace0c0","source":{"kind":"arxiv","id":"2204.09179","version":3},"attestation_state":"computed","paper":{"title":"On the Representation Collapse of Sparse Mixture of Experts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Barun Patra, Damai Dai, Furu Wei, Heyan Huang, Li Dong, Payal Bajaj, Saksham Singhal, Shaohan Huang, Shuming Ma, Xian-Ling Mao, Xia Song, Zewen Chi","submitted_at":"2022-04-20T01:40:19Z","abstract_excerpt":"Sparse mixture of experts provides larger model capacity while requiring a constant computational overhead. It employs the routing mechanism to distribute input tokens to the best-matched experts according to their hidden representations. However, learning such a routing mechanism encourages token clustering around expert centroids, implying a trend toward representation collapse. In this work, we propose to estimate the routing scores between tokens and experts on a low-dimensional hypersphere. We conduct extensive experiments on cross-lingual language model pre-training and fine-tuning on do"},"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":"2204.09179","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-04-20T01:40:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"1a7b42cb04ba72572f7cf0e315a49e0588fe67185402f1b084ab1b6eca0b2a43","abstract_canon_sha256":"b65dc290e17b6f6160533e7f0a94c8b29faafacfd60b8d3ca52fdc4e11743601"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:05:46.669046Z","signature_b64":"Y050R1FOs6oH6KvWZTd7YkgXNxj6oID/6I7Env8hYJ+HWGPzx06//g+IobiVkJw8TrCJg9RxK+hZyiUX6b0oCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"499226ebef63471aff959adda59cf5ce58ab4090da69114215f70c44f2ace0c0","last_reissued_at":"2026-07-05T05:05:46.668565Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:05:46.668565Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Representation Collapse of Sparse Mixture of Experts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Barun Patra, Damai Dai, Furu Wei, Heyan Huang, Li Dong, Payal Bajaj, Saksham Singhal, Shaohan Huang, Shuming Ma, Xian-Ling Mao, Xia Song, Zewen Chi","submitted_at":"2022-04-20T01:40:19Z","abstract_excerpt":"Sparse mixture of experts provides larger model capacity while requiring a constant computational overhead. It employs the routing mechanism to distribute input tokens to the best-matched experts according to their hidden representations. However, learning such a routing mechanism encourages token clustering around expert centroids, implying a trend toward representation collapse. In this work, we propose to estimate the routing scores between tokens and experts on a low-dimensional hypersphere. We conduct extensive experiments on cross-lingual language model pre-training and fine-tuning on do"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2204.09179","kind":"arxiv","version":3},"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/2204.09179/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":"2204.09179","created_at":"2026-07-05T05:05:46.668623+00:00"},{"alias_kind":"arxiv_version","alias_value":"2204.09179v3","created_at":"2026-07-05T05:05:46.668623+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2204.09179","created_at":"2026-07-05T05:05:46.668623+00:00"},{"alias_kind":"pith_short_12","alias_value":"JGJCN27PMNDR","created_at":"2026-07-05T05:05:46.668623+00:00"},{"alias_kind":"pith_short_16","alias_value":"JGJCN27PMNDRV74V","created_at":"2026-07-05T05:05:46.668623+00:00"},{"alias_kind":"pith_short_8","alias_value":"JGJCN27P","created_at":"2026-07-05T05:05:46.668623+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2412.00069","citing_title":"Condense, Don't Just Prune: Enhancing Efficiency and Performance in MoE Layer Pruning","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2502.05564","citing_title":"TabICL: A Tabular Foundation Model for In-Context Learning on Large Data","ref_index":90,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12476","citing_title":"Routers Learn the Geometry of Their Experts: Geometric Coupling in Sparse Mixture-of-Experts","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ","json":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ.json","graph_json":"https://pith.science/api/pith-number/JGJCN27PMNDRV74VTLO2LHHVZZ/graph.json","events_json":"https://pith.science/api/pith-number/JGJCN27PMNDRV74VTLO2LHHVZZ/events.json","paper":"https://pith.science/paper/JGJCN27P"},"agent_actions":{"view_html":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ","download_json":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ.json","view_paper":"https://pith.science/paper/JGJCN27P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2204.09179&json=true","fetch_graph":"https://pith.science/api/pith-number/JGJCN27PMNDRV74VTLO2LHHVZZ/graph.json","fetch_events":"https://pith.science/api/pith-number/JGJCN27PMNDRV74VTLO2LHHVZZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ/action/storage_attestation","attest_author":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ/action/author_attestation","sign_citation":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ/action/citation_signature","submit_replication":"https://pith.science/pith/JGJCN27PMNDRV74VTLO2LHHVZZ/action/replication_record"}},"created_at":"2026-07-05T05:05:46.668623+00:00","updated_at":"2026-07-05T05:05:46.668623+00:00"}