{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NTTZIGJ6M4MVTFCQKJWEICYZLM","short_pith_number":"pith:NTTZIGJ6","schema_version":"1.0","canonical_sha256":"6ce794193e6719599450526c440b195b15a2e304abd849e28626ac29275ceabb","source":{"kind":"arxiv","id":"2402.10790","version":2},"attestation_state":"computed","paper":{"title":"In Search of Needles in a 11M Haystack: Recurrent Memory Finds What LLMs Miss","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Artyom Sorokin, Aydar Bulatov, Dmitry Sorokin, Mikhail Burtsev, Petr Anokhin, Yuri Kuratov","submitted_at":"2024-02-16T16:15:01Z","abstract_excerpt":"This paper addresses the challenge of processing long documents using generative transformer models. To evaluate different approaches, we introduce BABILong, a new benchmark designed to assess model capabilities in extracting and processing distributed facts within extensive texts. Our evaluation, which includes benchmarks for GPT-4 and RAG, reveals that common methods are effective only for sequences up to $10^4$ elements. In contrast, fine-tuning GPT-2 with recurrent memory augmentations enables it to handle tasks involving up to $11\\times 10^6$ elements. This achievement marks a substantial"},"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":"2402.10790","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-16T16:15:01Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"b5e78fd8c693d9bbea30f6545fdeffcda7002c2faed05f2a61fb94f366160f10","abstract_canon_sha256":"764c5dd5462fd5496619cfc8e1da6676cd4e61fd58fcbee3c4c76881da518f95"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:47:29.020336Z","signature_b64":"nUz8TXNjRG6SxljoRSZKbiELetGj51NEkzLL0qqsDWmrhzx9xsfNhgxgyBWLNOnJkiPivXlHm8EJrZPqmCQVCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6ce794193e6719599450526c440b195b15a2e304abd849e28626ac29275ceabb","last_reissued_at":"2026-07-05T07:47:29.019834Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:47:29.019834Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"In Search of Needles in a 11M Haystack: Recurrent Memory Finds What LLMs Miss","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Artyom Sorokin, Aydar Bulatov, Dmitry Sorokin, Mikhail Burtsev, Petr Anokhin, Yuri Kuratov","submitted_at":"2024-02-16T16:15:01Z","abstract_excerpt":"This paper addresses the challenge of processing long documents using generative transformer models. To evaluate different approaches, we introduce BABILong, a new benchmark designed to assess model capabilities in extracting and processing distributed facts within extensive texts. Our evaluation, which includes benchmarks for GPT-4 and RAG, reveals that common methods are effective only for sequences up to $10^4$ elements. In contrast, fine-tuning GPT-2 with recurrent memory augmentations enables it to handle tasks involving up to $11\\times 10^6$ elements. This achievement marks a substantial"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.10790","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/2402.10790/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":"2402.10790","created_at":"2026-07-05T07:47:29.019891+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.10790v2","created_at":"2026-07-05T07:47:29.019891+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.10790","created_at":"2026-07-05T07:47:29.019891+00:00"},{"alias_kind":"pith_short_12","alias_value":"NTTZIGJ6M4MV","created_at":"2026-07-05T07:47:29.019891+00:00"},{"alias_kind":"pith_short_16","alias_value":"NTTZIGJ6M4MVTFCQ","created_at":"2026-07-05T07:47:29.019891+00:00"},{"alias_kind":"pith_short_8","alias_value":"NTTZIGJ6","created_at":"2026-07-05T07:47:29.019891+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22645","citing_title":"All Relations Lead to Rome: Automated Knowledge Graph Creation and Question Generation","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22645","citing_title":"All Relations Lead to Rome: Automated Knowledge Graph Creation and Question Generation","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30093","citing_title":"Efficient Retrieval-Augmented Generation via Token Co-occurrence Graphs","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2402.19473","citing_title":"Retrieval-Augmented Generation for AI-Generated Content: A Survey","ref_index":97,"is_internal_anchor":false},{"citing_arxiv_id":"2410.05779","citing_title":"LightRAG: Simple and Fast Retrieval-Augmented Generation","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM","json":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM.json","graph_json":"https://pith.science/api/pith-number/NTTZIGJ6M4MVTFCQKJWEICYZLM/graph.json","events_json":"https://pith.science/api/pith-number/NTTZIGJ6M4MVTFCQKJWEICYZLM/events.json","paper":"https://pith.science/paper/NTTZIGJ6"},"agent_actions":{"view_html":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM","download_json":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM.json","view_paper":"https://pith.science/paper/NTTZIGJ6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.10790&json=true","fetch_graph":"https://pith.science/api/pith-number/NTTZIGJ6M4MVTFCQKJWEICYZLM/graph.json","fetch_events":"https://pith.science/api/pith-number/NTTZIGJ6M4MVTFCQKJWEICYZLM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM/action/storage_attestation","attest_author":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM/action/author_attestation","sign_citation":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM/action/citation_signature","submit_replication":"https://pith.science/pith/NTTZIGJ6M4MVTFCQKJWEICYZLM/action/replication_record"}},"created_at":"2026-07-05T07:47:29.019891+00:00","updated_at":"2026-07-05T07:47:29.019891+00:00"}