{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:LHWNQN3FLUULJLMEOX4ONXODSS","short_pith_number":"pith:LHWNQN3F","schema_version":"1.0","canonical_sha256":"59ecd837655d28b4ad8475f8e6ddc394aabca4a8a0ce2f020acc61a96266dc8b","source":{"kind":"arxiv","id":"2608.12419","version":1},"attestation_state":"computed","paper":{"title":"LoKiFormer: Locality-aware Attention with Decoupled Knowledge Memory for Efficient Large Language Model Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Mingkui Tan, Qiuwu Chen, Ryan Dong, Simeng Ma, Yaofo Chen, Yifan Zhang, Ying Sun, Yuchen Li, Zeng You, Zhijie Qiu, Zimo Liu","submitted_at":"2026-08-12T07:45:11Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable breakthroughs across various applications. However, their architectures remain inefficient in pretraining due to two main limitations: (i) self-attention lacks an explicit inductive bias for locality, leading to redundant modeling of sequence-internal local information; (ii) mixture-of-experts (MoE) implicitly couples knowledge storage with computational pathways, hindering flexible access to sequence-external global knowledge. To overcome these limitations, we propose LoKiFormer, a novel LLM architecture that augments the standard decoder "},"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":"2608.12419","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-12T07:45:11Z","cross_cats_sorted":[],"title_canon_sha256":"897b757639f6f422d50e38b1f6850e6678c87a95cc9b0833b55201a19304ec85","abstract_canon_sha256":"c691f6960b56680b646c38212262af38bae1508c67390b8b09fdab54670cd0df"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-14T00:42:53.046924Z","signature_b64":"RQtwE0JtzMOIwJ0q0/YIJkFCFM9zJ95eUAITTpqO9w/Dt+aAxyBwqgEAj/2Fv6rvuOX7nQbiVFZnY0jBjNLnBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"59ecd837655d28b4ad8475f8e6ddc394aabca4a8a0ce2f020acc61a96266dc8b","last_reissued_at":"2026-08-14T00:42:53.025970Z","signature_status":"signed_v1","first_computed_at":"2026-08-14T00:42:53.025970Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoKiFormer: Locality-aware Attention with Decoupled Knowledge Memory for Efficient Large Language Model Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Mingkui Tan, Qiuwu Chen, Ryan Dong, Simeng Ma, Yaofo Chen, Yifan Zhang, Ying Sun, Yuchen Li, Zeng You, Zhijie Qiu, Zimo Liu","submitted_at":"2026-08-12T07:45:11Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable breakthroughs across various applications. However, their architectures remain inefficient in pretraining due to two main limitations: (i) self-attention lacks an explicit inductive bias for locality, leading to redundant modeling of sequence-internal local information; (ii) mixture-of-experts (MoE) implicitly couples knowledge storage with computational pathways, hindering flexible access to sequence-external global knowledge. To overcome these limitations, we propose LoKiFormer, a novel LLM architecture that augments the standard decoder "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.12419","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/2608.12419/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":"2608.12419","created_at":"2026-08-14T00:42:53.044847+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.12419v1","created_at":"2026-08-14T00:42:53.044847+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.12419","created_at":"2026-08-14T00:42:53.044847+00:00"},{"alias_kind":"pith_short_12","alias_value":"LHWNQN3FLUUL","created_at":"2026-08-14T00:42:53.044847+00:00"},{"alias_kind":"pith_short_16","alias_value":"LHWNQN3FLUULJLME","created_at":"2026-08-14T00:42:53.044847+00:00"},{"alias_kind":"pith_short_8","alias_value":"LHWNQN3F","created_at":"2026-08-14T00:42:53.044847+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/LHWNQN3FLUULJLMEOX4ONXODSS","json":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS.json","graph_json":"https://pith.science/api/pith-number/LHWNQN3FLUULJLMEOX4ONXODSS/graph.json","events_json":"https://pith.science/api/pith-number/LHWNQN3FLUULJLMEOX4ONXODSS/events.json","paper":"https://pith.science/paper/LHWNQN3F"},"agent_actions":{"view_html":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS","download_json":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS.json","view_paper":"https://pith.science/paper/LHWNQN3F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.12419&json=true","fetch_graph":"https://pith.science/api/pith-number/LHWNQN3FLUULJLMEOX4ONXODSS/graph.json","fetch_events":"https://pith.science/api/pith-number/LHWNQN3FLUULJLMEOX4ONXODSS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS/action/storage_attestation","attest_author":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS/action/author_attestation","sign_citation":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS/action/citation_signature","submit_replication":"https://pith.science/pith/LHWNQN3FLUULJLMEOX4ONXODSS/action/replication_record"}},"created_at":"2026-08-14T00:42:53.044847+00:00","updated_at":"2026-08-14T00:42:53.044847+00:00"}