{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:HGNUMGCMGAKCUNNONMRSIDPLRT","short_pith_number":"pith:HGNUMGCM","schema_version":"1.0","canonical_sha256":"399b46184c30142a35ae6b23240deb8cd99427a214e9f43522a80495dbdb9e05","source":{"kind":"arxiv","id":"2407.14482","version":3},"attestation_state":"computed","paper":{"title":"ChatQA 2: Bridging the Gap to Proprietary LLMs in Long Context and RAG Capabilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bryan Catanzaro, Chejian Xu, Mohammad Shoeybi, Peng Xu, Wei Ping, Xianchao Wu, Zihan Liu","submitted_at":"2024-07-19T17:35:47Z","abstract_excerpt":"In this work, we introduce ChatQA 2, an Llama 3.0-based model with a 128K context window, designed to bridge the gap between open-source LLMs and leading proprietary models (e.g., GPT-4-Turbo-2024-04-09) in long context understanding and retrieval-augmented generation (RAG) capabilities. These two capabilities are complementary to each other and essential for LLMs to process large volumes of information that cannot fit into a single prompt. We present a detailed continued training recipe to extend the context window of Llama3-70B-base from 8K to 128K tokens, along with a three-stage instructio"},"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":"2407.14482","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-19T17:35:47Z","cross_cats_sorted":["cs.AI","cs.IR","cs.LG"],"title_canon_sha256":"ac6b88c202a069653febaeaab925ac846bfc21f91b2f993dacd3f4f06f28863a","abstract_canon_sha256":"01f4b72751d433f8d6b040f9647bfa35bb8c71bc509e508ecc78303b0830016b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:14:43.668480Z","signature_b64":"6w8qX3cnyjbj3rzVwgB6yeG0KYEWmcU7XHoqrO6RdWQc+4uXxxYckQg4XdxIi79y2hX6XP8dqXquaItBZFQUBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"399b46184c30142a35ae6b23240deb8cd99427a214e9f43522a80495dbdb9e05","last_reissued_at":"2026-07-05T10:14:43.667948Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:14:43.667948Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ChatQA 2: Bridging the Gap to Proprietary LLMs in Long Context and RAG Capabilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bryan Catanzaro, Chejian Xu, Mohammad Shoeybi, Peng Xu, Wei Ping, Xianchao Wu, Zihan Liu","submitted_at":"2024-07-19T17:35:47Z","abstract_excerpt":"In this work, we introduce ChatQA 2, an Llama 3.0-based model with a 128K context window, designed to bridge the gap between open-source LLMs and leading proprietary models (e.g., GPT-4-Turbo-2024-04-09) in long context understanding and retrieval-augmented generation (RAG) capabilities. These two capabilities are complementary to each other and essential for LLMs to process large volumes of information that cannot fit into a single prompt. We present a detailed continued training recipe to extend the context window of Llama3-70B-base from 8K to 128K tokens, along with a three-stage instructio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.14482","kind":"arxiv","version":3},"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/2407.14482/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":"2407.14482","created_at":"2026-07-05T10:14:43.668009+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.14482v3","created_at":"2026-07-05T10:14:43.668009+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.14482","created_at":"2026-07-05T10:14:43.668009+00:00"},{"alias_kind":"pith_short_12","alias_value":"HGNUMGCMGAKC","created_at":"2026-07-05T10:14:43.668009+00:00"},{"alias_kind":"pith_short_16","alias_value":"HGNUMGCMGAKCUNNO","created_at":"2026-07-05T10:14:43.668009+00:00"},{"alias_kind":"pith_short_8","alias_value":"HGNUMGCM","created_at":"2026-07-05T10:14:43.668009+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11470","citing_title":"The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes","ref_index":277,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09659","citing_title":"End-to-End Context Compression at Scale","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2504.02181","citing_title":"A Survey of Scaling in Large Language Model Reasoning","ref_index":236,"is_internal_anchor":false},{"citing_arxiv_id":"2410.10594","citing_title":"VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07394","citing_title":"Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT","json":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT.json","graph_json":"https://pith.science/api/pith-number/HGNUMGCMGAKCUNNONMRSIDPLRT/graph.json","events_json":"https://pith.science/api/pith-number/HGNUMGCMGAKCUNNONMRSIDPLRT/events.json","paper":"https://pith.science/paper/HGNUMGCM"},"agent_actions":{"view_html":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT","download_json":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT.json","view_paper":"https://pith.science/paper/HGNUMGCM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.14482&json=true","fetch_graph":"https://pith.science/api/pith-number/HGNUMGCMGAKCUNNONMRSIDPLRT/graph.json","fetch_events":"https://pith.science/api/pith-number/HGNUMGCMGAKCUNNONMRSIDPLRT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT/action/storage_attestation","attest_author":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT/action/author_attestation","sign_citation":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT/action/citation_signature","submit_replication":"https://pith.science/pith/HGNUMGCMGAKCUNNONMRSIDPLRT/action/replication_record"}},"created_at":"2026-07-05T10:14:43.668009+00:00","updated_at":"2026-07-05T10:14:43.668009+00:00"}