{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:EI4V2FMTSLXIYEQHFWZYC7SAYN","short_pith_number":"pith:EI4V2FMT","schema_version":"1.0","canonical_sha256":"22395d159392ee8c12072db3817e40c35dc0c46922b445d6e66d369d977c6304","source":{"kind":"arxiv","id":"2407.12854","version":1},"attestation_state":"computed","paper":{"title":"Scaling Retrieval-Based Language Models with a Trillion-Token Datastore","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Akari Asai, Jacqueline He, Luke Zettlemoyer, Pang Wei Koh, Rulin Shao, Sewon Min, Tim Dettmers, Weijia Shi","submitted_at":"2024-07-09T08:27:27Z","abstract_excerpt":"Scaling laws with respect to the amount of training data and the number of parameters allow us to predict the cost-benefit trade-offs of pretraining language models (LMs) in different configurations. In this paper, we consider another dimension of scaling: the amount of data available at inference time. Specifically, we find that increasing the size of the datastore used by a retrieval-based LM monotonically improves language modeling and several downstream tasks without obvious saturation, such that a smaller model augmented with a large datastore outperforms a larger LM-only model on knowled"},"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":"2407.12854","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-09T08:27:27Z","cross_cats_sorted":["cs.AI","cs.IR","cs.LG"],"title_canon_sha256":"b2026cea641933523da41a3f245419298ac409758bcd57d5de305bc19a2da68e","abstract_canon_sha256":"33472807aa0272b4c8c2bb215b64589cf959ee09c53081a9f69a79db9a13827e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:45:12.636392Z","signature_b64":"HM2QjybIbNcQV37OIduOTrKdWDQ0RvuYpK95rnWpXxztSDkfAjdfQwjQWIJi4tIQy8FfXzBxe0afIK4N1XyFAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"22395d159392ee8c12072db3817e40c35dc0c46922b445d6e66d369d977c6304","last_reissued_at":"2026-07-05T08:45:12.635907Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:45:12.635907Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Retrieval-Based Language Models with a Trillion-Token Datastore","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Akari Asai, Jacqueline He, Luke Zettlemoyer, Pang Wei Koh, Rulin Shao, Sewon Min, Tim Dettmers, Weijia Shi","submitted_at":"2024-07-09T08:27:27Z","abstract_excerpt":"Scaling laws with respect to the amount of training data and the number of parameters allow us to predict the cost-benefit trade-offs of pretraining language models (LMs) in different configurations. In this paper, we consider another dimension of scaling: the amount of data available at inference time. Specifically, we find that increasing the size of the datastore used by a retrieval-based LM monotonically improves language modeling and several downstream tasks without obvious saturation, such that a smaller model augmented with a large datastore outperforms a larger LM-only model on knowled"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.12854","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/2407.12854/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":"2407.12854","created_at":"2026-07-05T08:45:12.635983+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.12854v1","created_at":"2026-07-05T08:45:12.635983+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.12854","created_at":"2026-07-05T08:45:12.635983+00:00"},{"alias_kind":"pith_short_12","alias_value":"EI4V2FMTSLXI","created_at":"2026-07-05T08:45:12.635983+00:00"},{"alias_kind":"pith_short_16","alias_value":"EI4V2FMTSLXIYEQH","created_at":"2026-07-05T08:45:12.635983+00:00"},{"alias_kind":"pith_short_8","alias_value":"EI4V2FMT","created_at":"2026-07-05T08:45:12.635983+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06641","citing_title":"Healthier LLMs: Retrieval-Augmented Generation for Public Health Question Answering","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2606.30518","citing_title":"Regime-Aware Peer Specialization for Robust RAG under Heterogeneous Knowledge Conflicts","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2603.04816","citing_title":"Scaling Laws for Cross-Encoder Reranking","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2407.21787","citing_title":"Large Language Monkeys: Scaling Inference Compute with Repeated Sampling","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN","json":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN.json","graph_json":"https://pith.science/api/pith-number/EI4V2FMTSLXIYEQHFWZYC7SAYN/graph.json","events_json":"https://pith.science/api/pith-number/EI4V2FMTSLXIYEQHFWZYC7SAYN/events.json","paper":"https://pith.science/paper/EI4V2FMT"},"agent_actions":{"view_html":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN","download_json":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN.json","view_paper":"https://pith.science/paper/EI4V2FMT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.12854&json=true","fetch_graph":"https://pith.science/api/pith-number/EI4V2FMTSLXIYEQHFWZYC7SAYN/graph.json","fetch_events":"https://pith.science/api/pith-number/EI4V2FMTSLXIYEQHFWZYC7SAYN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN/action/storage_attestation","attest_author":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN/action/author_attestation","sign_citation":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN/action/citation_signature","submit_replication":"https://pith.science/pith/EI4V2FMTSLXIYEQHFWZYC7SAYN/action/replication_record"}},"created_at":"2026-07-05T08:45:12.635983+00:00","updated_at":"2026-07-05T08:45:12.635983+00:00"}