{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MK3FHH6BVOUH5CFIBRB3GQ4H6T","short_pith_number":"pith:MK3FHH6B","schema_version":"1.0","canonical_sha256":"62b6539fc1aba87e88a80c43b34387f4f392a9a5a6ecb0f844f3579932d6983f","source":{"kind":"arxiv","id":"2512.13996","version":2},"attestation_state":"computed","paper":{"title":"DTop-p MoE: Sparsity-Controlled Dynamic Top-p MoE for Foundation Model Pre-training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Amit Hasan, Can Jin, Dimitris N. Metaxas, Hongwu Peng, Mingcan Xiang, Ohiremen Dibua, Qixin Zhang, Xiangchi Yuan, Yan Kang, Yifan Gong","submitted_at":"2025-12-16T01:28:57Z","abstract_excerpt":"Sparse Mixture-of-Experts architectures are essential for scaling model capacity efficiently, yet the standard Top-$k$ routing imposes a rigid sparsity pattern that ignores the intrinsic variance in token difficulty and layer-specific computational needs. Top-$p$ routing is more adaptive because it selects experts until their cumulative routing probability reaches a threshold, allowing confident tokens to use fewer experts and ambiguous tokens to recruit more. However, we demonstrate that existing naive Top-$p$ implementations with fixed global probability thresholds provide only marginal gain"},"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":"2512.13996","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-12-16T01:28:57Z","cross_cats_sorted":[],"title_canon_sha256":"247ee98c27a8d3b1b7f7ea4be7ece654ef24f179b76dacbe5cbb837bcd75ca00","abstract_canon_sha256":"bbf8e4ad132b20cce5dc96972a876fb8d2b6f5bc52e0db7e92ad0ea0411276c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T02:03:31.682992Z","signature_b64":"BskZ4KtLrSLMnzsylRzWYp4zU4/VSDprLWhJ8pgXMekL4aAhgBNN7z8nWIHu9pRUsnU2Vt2CBfpaJw30qZM9Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"62b6539fc1aba87e88a80c43b34387f4f392a9a5a6ecb0f844f3579932d6983f","last_reissued_at":"2026-06-01T02:03:31.681817Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T02:03:31.681817Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DTop-p MoE: Sparsity-Controlled Dynamic Top-p MoE for Foundation Model Pre-training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Amit Hasan, Can Jin, Dimitris N. Metaxas, Hongwu Peng, Mingcan Xiang, Ohiremen Dibua, Qixin Zhang, Xiangchi Yuan, Yan Kang, Yifan Gong","submitted_at":"2025-12-16T01:28:57Z","abstract_excerpt":"Sparse Mixture-of-Experts architectures are essential for scaling model capacity efficiently, yet the standard Top-$k$ routing imposes a rigid sparsity pattern that ignores the intrinsic variance in token difficulty and layer-specific computational needs. Top-$p$ routing is more adaptive because it selects experts until their cumulative routing probability reaches a threshold, allowing confident tokens to use fewer experts and ambiguous tokens to recruit more. However, we demonstrate that existing naive Top-$p$ implementations with fixed global probability thresholds provide only marginal gain"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2512.13996","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/2512.13996/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":"2512.13996","created_at":"2026-06-01T02:03:31.681958+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.13996v2","created_at":"2026-06-01T02:03:31.681958+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.13996","created_at":"2026-06-01T02:03:31.681958+00:00"},{"alias_kind":"pith_short_12","alias_value":"MK3FHH6BVOUH","created_at":"2026-06-01T02:03:31.681958+00:00"},{"alias_kind":"pith_short_16","alias_value":"MK3FHH6BVOUH5CFI","created_at":"2026-06-01T02:03:31.681958+00:00"},{"alias_kind":"pith_short_8","alias_value":"MK3FHH6B","created_at":"2026-06-01T02:03:31.681958+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.14438","citing_title":"BEAM: Binary Expert Activation Masking for Dynamic Routing in MoE","ref_index":11,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T","json":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T.json","graph_json":"https://pith.science/api/pith-number/MK3FHH6BVOUH5CFIBRB3GQ4H6T/graph.json","events_json":"https://pith.science/api/pith-number/MK3FHH6BVOUH5CFIBRB3GQ4H6T/events.json","paper":"https://pith.science/paper/MK3FHH6B"},"agent_actions":{"view_html":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T","download_json":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T.json","view_paper":"https://pith.science/paper/MK3FHH6B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.13996&json=true","fetch_graph":"https://pith.science/api/pith-number/MK3FHH6BVOUH5CFIBRB3GQ4H6T/graph.json","fetch_events":"https://pith.science/api/pith-number/MK3FHH6BVOUH5CFIBRB3GQ4H6T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T/action/storage_attestation","attest_author":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T/action/author_attestation","sign_citation":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T/action/citation_signature","submit_replication":"https://pith.science/pith/MK3FHH6BVOUH5CFIBRB3GQ4H6T/action/replication_record"}},"created_at":"2026-06-01T02:03:31.681958+00:00","updated_at":"2026-06-01T02:03:31.681958+00:00"}