{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:VUBIJUV72M6JWYNYG7PCEAHDOF","short_pith_number":"pith:VUBIJUV7","schema_version":"1.0","canonical_sha256":"ad0284d2bfd33c9b61b837de2200e3715b15baf37134c1f141948db519b79662","source":{"kind":"arxiv","id":"2402.18668","version":2},"attestation_state":"computed","paper":{"title":"Simple linear attention language models balance the recall-throughput tradeoff","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aman Timalsina, Atri Rudra, Christopher R\\'e, Dylan Zinsley, James Zou, Michael Zhang, Sabri Eyuboglu, Silas Alberti, Simran Arora","submitted_at":"2024-02-28T19:28:27Z","abstract_excerpt":"Recent work has shown that attention-based language models excel at recall, the ability to ground generations in tokens previously seen in context. However, the efficiency of attention-based models is bottle-necked during inference by the KV-cache's aggressive memory consumption. In this work, we explore whether we can improve language model efficiency (e.g. by reducing memory consumption) without compromising on recall. By applying experiments and theory to a broad set of architectures, we identify a key tradeoff between a model's state size and recall ability. We show that efficient alternat"},"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":"2402.18668","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-28T19:28:27Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"68d57bcc9a8e5ab91a31e1195629a917ae00683820b5c2ad5992c992a1420246","abstract_canon_sha256":"2972be310ad45758f91d09af4ce498bf258e980879e253fd8490b9bacf5413b7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:26:01.545862Z","signature_b64":"tZ11U9go5CsW5ZHOOY6GT33VBVYJn5xVvCmSpZsjnG+YvPQ2MvNZwj/WdluXg4LG5F5Yne0+r0FFjnbv7GhDBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ad0284d2bfd33c9b61b837de2200e3715b15baf37134c1f141948db519b79662","last_reissued_at":"2026-07-05T10:26:01.544747Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:26:01.544747Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Simple linear attention language models balance the recall-throughput tradeoff","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aman Timalsina, Atri Rudra, Christopher R\\'e, Dylan Zinsley, James Zou, Michael Zhang, Sabri Eyuboglu, Silas Alberti, Simran Arora","submitted_at":"2024-02-28T19:28:27Z","abstract_excerpt":"Recent work has shown that attention-based language models excel at recall, the ability to ground generations in tokens previously seen in context. However, the efficiency of attention-based models is bottle-necked during inference by the KV-cache's aggressive memory consumption. In this work, we explore whether we can improve language model efficiency (e.g. by reducing memory consumption) without compromising on recall. By applying experiments and theory to a broad set of architectures, we identify a key tradeoff between a model's state size and recall ability. We show that efficient alternat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.18668","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/2402.18668/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":"2402.18668","created_at":"2026-07-05T10:26:01.544864+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.18668v2","created_at":"2026-07-05T10:26:01.544864+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.18668","created_at":"2026-07-05T10:26:01.544864+00:00"},{"alias_kind":"pith_short_12","alias_value":"VUBIJUV72M6J","created_at":"2026-07-05T10:26:01.544864+00:00"},{"alias_kind":"pith_short_16","alias_value":"VUBIJUV72M6JWYNY","created_at":"2026-07-05T10:26:01.544864+00:00"},{"alias_kind":"pith_short_8","alias_value":"VUBIJUV7","created_at":"2026-07-05T10:26:01.544864+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":17,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07386","citing_title":"Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity","ref_index":80,"is_internal_anchor":true},{"citing_arxiv_id":"2607.02303","citing_title":"A Hippocampus for Linear Attention: An Exact Memory for What the Recurrent State Forgets","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09862","citing_title":"Blurry Window Attention","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30562","citing_title":"Morphing into Hybrid Attention Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26099","citing_title":"Do Language Models Need Sleep? Offline Recurrence for Improved Online Inference","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2506.13183","citing_title":"MT-PCR: Hybrid Mamba-Transformer Network with Spatial Serialization for Point Cloud Registration","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2507.09025","citing_title":"Lizard: An Efficient Linearization Framework for Large Language Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24552","citing_title":"Short window attention enables long-term memorization","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2505.23884","citing_title":"Test-Time Training Done Right","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2312.06635","citing_title":"Gated Linear Attention Transformers with Hardware-Efficient Training","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13473","citing_title":"OSDN: Improving Delta Rule with Provable Online Preconditioning in Linear Attention","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2412.06464","citing_title":"Gated Delta Networks: Improving Mamba2 with Delta Rule","ref_index":296,"is_internal_anchor":false},{"citing_arxiv_id":"2506.13585","citing_title":"MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05066","citing_title":"The Impossibility Triangle of Long-Context Modeling","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10064","citing_title":"On The Application of Linear Attention in Multimodal Transformers","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06946","citing_title":"Adaptive Memory Decay for Log-Linear Attention","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05030","citing_title":"Phase-Associative Memory: Sequence Modeling in Complex Hilbert Space","ref_index":116,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF","json":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF.json","graph_json":"https://pith.science/api/pith-number/VUBIJUV72M6JWYNYG7PCEAHDOF/graph.json","events_json":"https://pith.science/api/pith-number/VUBIJUV72M6JWYNYG7PCEAHDOF/events.json","paper":"https://pith.science/paper/VUBIJUV7"},"agent_actions":{"view_html":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF","download_json":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF.json","view_paper":"https://pith.science/paper/VUBIJUV7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.18668&json=true","fetch_graph":"https://pith.science/api/pith-number/VUBIJUV72M6JWYNYG7PCEAHDOF/graph.json","fetch_events":"https://pith.science/api/pith-number/VUBIJUV72M6JWYNYG7PCEAHDOF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF/action/storage_attestation","attest_author":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF/action/author_attestation","sign_citation":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF/action/citation_signature","submit_replication":"https://pith.science/pith/VUBIJUV72M6JWYNYG7PCEAHDOF/action/replication_record"}},"created_at":"2026-07-05T10:26:01.544864+00:00","updated_at":"2026-07-05T10:26:01.544864+00:00"}