{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VOAAFGPV2AR5FWC76PXSPY3YY4","short_pith_number":"pith:VOAAFGPV","schema_version":"1.0","canonical_sha256":"ab800299f5d023d2d85ff3ef27e378c7280853a627a619a1559985ef0e6e846b","source":{"kind":"arxiv","id":"2504.12637","version":1},"attestation_state":"computed","paper":{"title":"Scaling Instruction-Tuned LLMs to Million-Token Contexts via Hierarchical Synthetic Data Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ben Athiwaratkun, Ce Zhang, Jue Wang, Linda He, Maurice Weber, Shang Zhu","submitted_at":"2025-04-17T04:46:57Z","abstract_excerpt":"Large Language Models (LLMs) struggle with long-context reasoning, not only due to the quadratic scaling of computational complexity with sequence length but also because of the scarcity and expense of annotating long-context data. There has been barely any open-source work that systematically ablates long-context data, nor is there any openly available instruction tuning dataset with contexts surpassing 100K tokens. To bridge this gap, we introduce a novel post-training synthetic data generation strategy designed to efficiently extend the context window of LLMs while preserving their general "},"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":"2504.12637","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-17T04:46:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1b99cceb520d3b249f23fc9eff900c68844c57e0b44316a3de6b97a62777157c","abstract_canon_sha256":"de179f81b55dfea093520650e0068a725f840f9fb1a26a1b6fb525d45ba574b8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:19.990918Z","signature_b64":"izvmuE7KQODpfQm1zt07REWY8xb80fTmTTmOonpv1ItDxLgGmXXYJNmR2toBW/BEjduF71Dn6EnqQUiYQ1UqAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ab800299f5d023d2d85ff3ef27e378c7280853a627a619a1559985ef0e6e846b","last_reissued_at":"2026-07-05T10:50:19.990405Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:19.990405Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Instruction-Tuned LLMs to Million-Token Contexts via Hierarchical Synthetic Data Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ben Athiwaratkun, Ce Zhang, Jue Wang, Linda He, Maurice Weber, Shang Zhu","submitted_at":"2025-04-17T04:46:57Z","abstract_excerpt":"Large Language Models (LLMs) struggle with long-context reasoning, not only due to the quadratic scaling of computational complexity with sequence length but also because of the scarcity and expense of annotating long-context data. There has been barely any open-source work that systematically ablates long-context data, nor is there any openly available instruction tuning dataset with contexts surpassing 100K tokens. To bridge this gap, we introduce a novel post-training synthetic data generation strategy designed to efficiently extend the context window of LLMs while preserving their general "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.12637","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/2504.12637/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":"2504.12637","created_at":"2026-07-05T10:50:19.990463+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.12637v1","created_at":"2026-07-05T10:50:19.990463+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.12637","created_at":"2026-07-05T10:50:19.990463+00:00"},{"alias_kind":"pith_short_12","alias_value":"VOAAFGPV2AR5","created_at":"2026-07-05T10:50:19.990463+00:00"},{"alias_kind":"pith_short_16","alias_value":"VOAAFGPV2AR5FWC7","created_at":"2026-07-05T10:50:19.990463+00:00"},{"alias_kind":"pith_short_8","alias_value":"VOAAFGPV","created_at":"2026-07-05T10:50:19.990463+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09659","citing_title":"End-to-End Context Compression at Scale","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4","json":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4.json","graph_json":"https://pith.science/api/pith-number/VOAAFGPV2AR5FWC76PXSPY3YY4/graph.json","events_json":"https://pith.science/api/pith-number/VOAAFGPV2AR5FWC76PXSPY3YY4/events.json","paper":"https://pith.science/paper/VOAAFGPV"},"agent_actions":{"view_html":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4","download_json":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4.json","view_paper":"https://pith.science/paper/VOAAFGPV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.12637&json=true","fetch_graph":"https://pith.science/api/pith-number/VOAAFGPV2AR5FWC76PXSPY3YY4/graph.json","fetch_events":"https://pith.science/api/pith-number/VOAAFGPV2AR5FWC76PXSPY3YY4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4/action/storage_attestation","attest_author":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4/action/author_attestation","sign_citation":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4/action/citation_signature","submit_replication":"https://pith.science/pith/VOAAFGPV2AR5FWC76PXSPY3YY4/action/replication_record"}},"created_at":"2026-07-05T10:50:19.990463+00:00","updated_at":"2026-07-05T10:50:19.990463+00:00"}