{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:U23BDIT7CV7A7GPYTKJEGUXY6B","short_pith_number":"pith:U23BDIT7","schema_version":"1.0","canonical_sha256":"a6b611a27f157e0f99f89a924352f8f06d2e4a13d07c2e21373a24b2604217fb","source":{"kind":"arxiv","id":"2410.01651","version":4},"attestation_state":"computed","paper":{"title":"Efficient Length-Generalizable Attention via Causal Retrieval for Long-Context Language Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jun Zhao, Kewei Tu, Wei Wu, Xiang Hu, Zhihao Teng","submitted_at":"2024-10-02T15:18:34Z","abstract_excerpt":"Despite the success of Transformers, handling long contexts remains challenging due to the limited length generalization and quadratic complexity of self-attention. Thus Transformers often require post-training with a larger attention window, significantly increasing computational and memory costs. In this paper, we propose a novel attention mechanism based on dynamic context, Grouped Cross Attention (GCA), which can generalize to 1000 times the pre-training context length while maintaining the ability to access distant information with a constant attention window size. For a given input seque"},"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":"2410.01651","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-02T15:18:34Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"403ae5b82b909cf15c0cb540cc18026b9b2b33181abf7fd9e450ddaff4e51aae","abstract_canon_sha256":"8ffead4ae20c6aad84a5fefbb8bc74758841cce886a97292eda2b4fcb5f79036"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:05.331934Z","signature_b64":"M96Th6KL/NnCPxJFVAPdWapB8ZS+ZZyNOnt/x0UcviVKI2HmxSE3kBk6nUkib8OzO/lZ14jMS892J9IbBTybCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a6b611a27f157e0f99f89a924352f8f06d2e4a13d07c2e21373a24b2604217fb","last_reissued_at":"2026-07-05T11:20:05.331369Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:05.331369Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Length-Generalizable Attention via Causal Retrieval for Long-Context Language Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jun Zhao, Kewei Tu, Wei Wu, Xiang Hu, Zhihao Teng","submitted_at":"2024-10-02T15:18:34Z","abstract_excerpt":"Despite the success of Transformers, handling long contexts remains challenging due to the limited length generalization and quadratic complexity of self-attention. Thus Transformers often require post-training with a larger attention window, significantly increasing computational and memory costs. In this paper, we propose a novel attention mechanism based on dynamic context, Grouped Cross Attention (GCA), which can generalize to 1000 times the pre-training context length while maintaining the ability to access distant information with a constant attention window size. For a given input seque"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.01651","kind":"arxiv","version":4},"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/2410.01651/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":"2410.01651","created_at":"2026-07-05T11:20:05.331452+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.01651v4","created_at":"2026-07-05T11:20:05.331452+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.01651","created_at":"2026-07-05T11:20:05.331452+00:00"},{"alias_kind":"pith_short_12","alias_value":"U23BDIT7CV7A","created_at":"2026-07-05T11:20:05.331452+00:00"},{"alias_kind":"pith_short_16","alias_value":"U23BDIT7CV7A7GPY","created_at":"2026-07-05T11:20:05.331452+00:00"},{"alias_kind":"pith_short_8","alias_value":"U23BDIT7","created_at":"2026-07-05T11:20:05.331452+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01538","citing_title":"Can Language Models Actually Retrieve In-Context? Drowning in Documents at Million Token Scale","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B","json":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B.json","graph_json":"https://pith.science/api/pith-number/U23BDIT7CV7A7GPYTKJEGUXY6B/graph.json","events_json":"https://pith.science/api/pith-number/U23BDIT7CV7A7GPYTKJEGUXY6B/events.json","paper":"https://pith.science/paper/U23BDIT7"},"agent_actions":{"view_html":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B","download_json":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B.json","view_paper":"https://pith.science/paper/U23BDIT7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.01651&json=true","fetch_graph":"https://pith.science/api/pith-number/U23BDIT7CV7A7GPYTKJEGUXY6B/graph.json","fetch_events":"https://pith.science/api/pith-number/U23BDIT7CV7A7GPYTKJEGUXY6B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B/action/storage_attestation","attest_author":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B/action/author_attestation","sign_citation":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B/action/citation_signature","submit_replication":"https://pith.science/pith/U23BDIT7CV7A7GPYTKJEGUXY6B/action/replication_record"}},"created_at":"2026-07-05T11:20:05.331452+00:00","updated_at":"2026-07-05T11:20:05.331452+00:00"}