{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4VZEGWFNDWUHB6UOIDHC7WMNSR","short_pith_number":"pith:4VZEGWFN","schema_version":"1.0","canonical_sha256":"e5724358ad1da870fa8e40ce2fd98d945f07103460809f808417678f178ef8fe","source":{"kind":"arxiv","id":"2506.13541","version":1},"attestation_state":"computed","paper":{"title":"Mixture of Weight-shared Heterogeneous Group Attention Experts for Dynamic Token-wise KV Optimization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Cheng-zhong Xu, Dongping Liao, Guanghui Song, Kejiang Ye, Xitong Gao, Yiren Zhao","submitted_at":"2025-06-16T14:30:17Z","abstract_excerpt":"Transformer models face scalability challenges in causal language modeling (CLM) due to inefficient memory allocation for growing key-value (KV) caches, which strains compute and storage resources. Existing methods like Grouped Query Attention (GQA) and token-level KV optimization improve efficiency but rely on rigid resource allocation, often discarding \"low-priority\" tokens or statically grouping them, failing to address the dynamic spectrum of token importance. We propose mixSGA, a novel mixture-of-expert (MoE) approach that dynamically optimizes token-wise computation and memory allocation"},"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":"2506.13541","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-16T14:30:17Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"8a545aeab97604ee024d66cb88de5a7b9fa0f8c579996e327a8152f5209e79d2","abstract_canon_sha256":"952ed779210cf0b63a28c32fc4238dacd1bb8a751b98ab5760313984011a7d8d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:22:21.896005Z","signature_b64":"mYGjyilxSag7kvETAx6t8oy/nD6+qhyUnBXzlMzrMCBCOweDG+gR29XkU5M0XllZi6OQU1sUNPB4/CqKuYg9Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e5724358ad1da870fa8e40ce2fd98d945f07103460809f808417678f178ef8fe","last_reissued_at":"2026-07-05T11:22:21.895509Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:22:21.895509Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mixture of Weight-shared Heterogeneous Group Attention Experts for Dynamic Token-wise KV Optimization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Cheng-zhong Xu, Dongping Liao, Guanghui Song, Kejiang Ye, Xitong Gao, Yiren Zhao","submitted_at":"2025-06-16T14:30:17Z","abstract_excerpt":"Transformer models face scalability challenges in causal language modeling (CLM) due to inefficient memory allocation for growing key-value (KV) caches, which strains compute and storage resources. Existing methods like Grouped Query Attention (GQA) and token-level KV optimization improve efficiency but rely on rigid resource allocation, often discarding \"low-priority\" tokens or statically grouping them, failing to address the dynamic spectrum of token importance. We propose mixSGA, a novel mixture-of-expert (MoE) approach that dynamically optimizes token-wise computation and memory allocation"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.13541","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/2506.13541/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":"2506.13541","created_at":"2026-07-05T11:22:21.895568+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.13541v1","created_at":"2026-07-05T11:22:21.895568+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.13541","created_at":"2026-07-05T11:22:21.895568+00:00"},{"alias_kind":"pith_short_12","alias_value":"4VZEGWFNDWUH","created_at":"2026-07-05T11:22:21.895568+00:00"},{"alias_kind":"pith_short_16","alias_value":"4VZEGWFNDWUHB6UO","created_at":"2026-07-05T11:22:21.895568+00:00"},{"alias_kind":"pith_short_8","alias_value":"4VZEGWFN","created_at":"2026-07-05T11:22:21.895568+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/4VZEGWFNDWUHB6UOIDHC7WMNSR","json":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR.json","graph_json":"https://pith.science/api/pith-number/4VZEGWFNDWUHB6UOIDHC7WMNSR/graph.json","events_json":"https://pith.science/api/pith-number/4VZEGWFNDWUHB6UOIDHC7WMNSR/events.json","paper":"https://pith.science/paper/4VZEGWFN"},"agent_actions":{"view_html":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR","download_json":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR.json","view_paper":"https://pith.science/paper/4VZEGWFN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.13541&json=true","fetch_graph":"https://pith.science/api/pith-number/4VZEGWFNDWUHB6UOIDHC7WMNSR/graph.json","fetch_events":"https://pith.science/api/pith-number/4VZEGWFNDWUHB6UOIDHC7WMNSR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR/action/storage_attestation","attest_author":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR/action/author_attestation","sign_citation":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR/action/citation_signature","submit_replication":"https://pith.science/pith/4VZEGWFNDWUHB6UOIDHC7WMNSR/action/replication_record"}},"created_at":"2026-07-05T11:22:21.895568+00:00","updated_at":"2026-07-05T11:22:21.895568+00:00"}