{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:S36NX22CQTY3POH4VRMEFUDSW5","short_pith_number":"pith:S36NX22C","schema_version":"1.0","canonical_sha256":"96fcdbeb4284f1b7b8fcac5842d072b7492c185367bfa8759fc7a86530ba4498","source":{"kind":"arxiv","id":"2506.23225","version":1},"attestation_state":"computed","paper":{"title":"Masked Gated Linear Unit","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ikuro Sato, Nakamasa Inoue, Rio Yokota, Yukito Tajima, Yusuke Sekikawa","submitted_at":"2025-06-29T13:16:20Z","abstract_excerpt":"Gated Linear Units (GLUs) have become essential components in the feed-forward networks of state-of-the-art Large Language Models (LLMs). However, they require twice as many memory reads compared to feed-forward layers without gating, due to the use of separate weight matrices for the gate and value streams. To address this bottleneck, we introduce Masked Gated Linear Units (MGLUs), a novel family of GLUs with an efficient kernel implementation. The core contribution of MGLUs include: (1) the Mixture of Element-wise Gating (MoEG) architecture that learns multiple binary masks, each determining"},"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":"2506.23225","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-29T13:16:20Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"73c18f22d6100cf3de88a8360c3207f8ce21aed9b61b488b53d8d93412f43f66","abstract_canon_sha256":"9165452d816091550ba9c4755ce73168e25b1fcdf5b488a6508a8657ea280e21"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:29:18.202822Z","signature_b64":"ogG2AhLMXNYaxkK+7PsEdI22DBeeVkEBVQm3JlgdHOiEScIRJWj9G0hdcnnTxgVP/hWZdPDiIws7jXlAaP+tCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"96fcdbeb4284f1b7b8fcac5842d072b7492c185367bfa8759fc7a86530ba4498","last_reissued_at":"2026-07-05T11:29:18.202341Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:29:18.202341Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Masked Gated Linear Unit","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ikuro Sato, Nakamasa Inoue, Rio Yokota, Yukito Tajima, Yusuke Sekikawa","submitted_at":"2025-06-29T13:16:20Z","abstract_excerpt":"Gated Linear Units (GLUs) have become essential components in the feed-forward networks of state-of-the-art Large Language Models (LLMs). However, they require twice as many memory reads compared to feed-forward layers without gating, due to the use of separate weight matrices for the gate and value streams. To address this bottleneck, we introduce Masked Gated Linear Units (MGLUs), a novel family of GLUs with an efficient kernel implementation. The core contribution of MGLUs include: (1) the Mixture of Element-wise Gating (MoEG) architecture that learns multiple binary masks, each determining"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.23225","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/2506.23225/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":"2506.23225","created_at":"2026-07-05T11:29:18.202399+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.23225v1","created_at":"2026-07-05T11:29:18.202399+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.23225","created_at":"2026-07-05T11:29:18.202399+00:00"},{"alias_kind":"pith_short_12","alias_value":"S36NX22CQTY3","created_at":"2026-07-05T11:29:18.202399+00:00"},{"alias_kind":"pith_short_16","alias_value":"S36NX22CQTY3POH4","created_at":"2026-07-05T11:29:18.202399+00:00"},{"alias_kind":"pith_short_8","alias_value":"S36NX22C","created_at":"2026-07-05T11:29:18.202399+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/S36NX22CQTY3POH4VRMEFUDSW5","json":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5.json","graph_json":"https://pith.science/api/pith-number/S36NX22CQTY3POH4VRMEFUDSW5/graph.json","events_json":"https://pith.science/api/pith-number/S36NX22CQTY3POH4VRMEFUDSW5/events.json","paper":"https://pith.science/paper/S36NX22C"},"agent_actions":{"view_html":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5","download_json":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5.json","view_paper":"https://pith.science/paper/S36NX22C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.23225&json=true","fetch_graph":"https://pith.science/api/pith-number/S36NX22CQTY3POH4VRMEFUDSW5/graph.json","fetch_events":"https://pith.science/api/pith-number/S36NX22CQTY3POH4VRMEFUDSW5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5/action/storage_attestation","attest_author":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5/action/author_attestation","sign_citation":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5/action/citation_signature","submit_replication":"https://pith.science/pith/S36NX22CQTY3POH4VRMEFUDSW5/action/replication_record"}},"created_at":"2026-07-05T11:29:18.202399+00:00","updated_at":"2026-07-05T11:29:18.202399+00:00"}