{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:KSIZCRJD2K6W5PIJ6GTXUJ6GV5","short_pith_number":"pith:KSIZCRJD","schema_version":"1.0","canonical_sha256":"5491914523d2bd6ebd09f1a77a27c6af76ba5c7b20e31e78e5c0d877e9af8f7d","source":{"kind":"arxiv","id":"2312.04927","version":1},"attestation_state":"computed","paper":{"title":"Zoology: Measuring and Improving Recall in Efficient Language Models","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, Isys Johnson, James Zou, Michael Poli, Sabri Eyuboglu, Simran Arora","submitted_at":"2023-12-08T09:44:25Z","abstract_excerpt":"Attention-free language models that combine gating and convolutions are growing in popularity due to their efficiency and increasingly competitive performance. To better understand these architectures, we pretrain a suite of 17 attention and \"gated-convolution\" language models, finding that SoTA gated-convolution architectures still underperform attention by up to 2.1 perplexity points on the Pile. In fine-grained analysis, we find 82% of the gap is explained by each model's ability to recall information that is previously mentioned in-context, e.g. \"Hakuna Matata means no worries Hakuna Matat"},"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":"2312.04927","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.CL","submitted_at":"2023-12-08T09:44:25Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d22562221fabf3f7b202b1ea2e4f1abbfd1b79a30c4dafded3c272fb8ef9e5b8","abstract_canon_sha256":"2cd007406634ab13a4857342668f582a28f872010a6e622384925fa49abbfc00"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:21:54.505476Z","signature_b64":"HgSMVm8fFkYJUG+11z+0xEfRN6CScts22cgxnf1/VHZDErN0CDBrINkOVwtI9eTefaaxPeHV5Jt9XHceBxwJCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5491914523d2bd6ebd09f1a77a27c6af76ba5c7b20e31e78e5c0d877e9af8f7d","last_reissued_at":"2026-07-05T07:21:54.505008Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:21:54.505008Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Zoology: Measuring and Improving Recall in Efficient Language Models","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, Isys Johnson, James Zou, Michael Poli, Sabri Eyuboglu, Simran Arora","submitted_at":"2023-12-08T09:44:25Z","abstract_excerpt":"Attention-free language models that combine gating and convolutions are growing in popularity due to their efficiency and increasingly competitive performance. To better understand these architectures, we pretrain a suite of 17 attention and \"gated-convolution\" language models, finding that SoTA gated-convolution architectures still underperform attention by up to 2.1 perplexity points on the Pile. In fine-grained analysis, we find 82% of the gap is explained by each model's ability to recall information that is previously mentioned in-context, e.g. \"Hakuna Matata means no worries Hakuna Matat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.04927","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/2312.04927/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":"2312.04927","created_at":"2026-07-05T07:21:54.505065+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.04927v1","created_at":"2026-07-05T07:21:54.505065+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.04927","created_at":"2026-07-05T07:21:54.505065+00:00"},{"alias_kind":"pith_short_12","alias_value":"KSIZCRJD2K6W","created_at":"2026-07-05T07:21:54.505065+00:00"},{"alias_kind":"pith_short_16","alias_value":"KSIZCRJD2K6W5PIJ","created_at":"2026-07-05T07:21:54.505065+00:00"},{"alias_kind":"pith_short_8","alias_value":"KSIZCRJD","created_at":"2026-07-05T07:21:54.505065+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":18,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.02303","citing_title":"A Hippocampus for Linear Attention: An Exact Memory for What the Recurrent State Forgets","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08804","citing_title":"Q-Delta: Beyond Key-Value Associative State Evolution","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03825","citing_title":"Dynamic Short Convolutions Improve Transformers","ref_index":155,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09862","citing_title":"Blurry Window Attention","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31163","citing_title":"Memory by Design: Probabilistic Sequence Layers","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29157","citing_title":"Parallax: Parameterized Local Linear Attention for Language Modeling","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2511.21016","citing_title":"Gated KalmaNet: A Fading Memory Layer Through Test-Time Ridge Regression","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2406.07887","citing_title":"An Empirical Study of Mamba-based Language Models","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2312.06635","citing_title":"Gated Linear Attention Transformers with Hardware-Efficient Training","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13473","citing_title":"OSDN: Improving Delta Rule with Provable Online Preconditioning in Linear Attention","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2510.26692","citing_title":"Kimi Linear: An Expressive, Efficient Attention Architecture","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2412.06464","citing_title":"Gated Delta Networks: Improving Mamba2 with Delta Rule","ref_index":294,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11196","citing_title":"Variational Linear Attention: Stable Associative Memory for Long-Context Transformers","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06554","citing_title":"Long Context Pre-Training with Lighthouse Attention","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22442","citing_title":"HubRouter: A Pluggable Sub-Quadratic Routing Primitive for Hybrid Sequence Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05838","citing_title":"MDN: Parallelizing Stepwise Momentum for Delta Linear Attention","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06946","citing_title":"Adaptive Memory Decay for Log-Linear Attention","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18580","citing_title":"Sessa: Selective State Space Attention","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5","json":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5.json","graph_json":"https://pith.science/api/pith-number/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/graph.json","events_json":"https://pith.science/api/pith-number/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/events.json","paper":"https://pith.science/paper/KSIZCRJD"},"agent_actions":{"view_html":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5","download_json":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5.json","view_paper":"https://pith.science/paper/KSIZCRJD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.04927&json=true","fetch_graph":"https://pith.science/api/pith-number/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/graph.json","fetch_events":"https://pith.science/api/pith-number/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/action/storage_attestation","attest_author":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/action/author_attestation","sign_citation":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/action/citation_signature","submit_replication":"https://pith.science/pith/KSIZCRJD2K6W5PIJ6GTXUJ6GV5/action/replication_record"}},"created_at":"2026-07-05T07:21:54.505065+00:00","updated_at":"2026-07-05T07:21:54.505065+00:00"}