{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KIMOEUGCQP4JSW46BQNO7WSQGS","short_pith_number":"pith:KIMOEUGC","schema_version":"1.0","canonical_sha256":"5218e250c283f8995b9e0c1aefda5034881cd1e3425295972246ca27e732c495","source":{"kind":"arxiv","id":"2502.20082","version":1},"attestation_state":"computed","paper":{"title":"LongRoPE2: Near-Lossless LLM Context Window Scaling","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fan Yang, Gaokai Zhang, Gilsinia Lopez, Li Lyna Zhang, Mao Yang, Ning Shang, Siyuan Wang, Weizhu Chen","submitted_at":"2025-02-27T13:41:07Z","abstract_excerpt":"LongRoPE2 is a novel approach that extends the effective context window of pre-trained large language models (LLMs) to the target length, while preserving the performance on the original shorter context window. This is achieved by three contributions: (1) a hypothesis that insufficient training in higher RoPE dimensions contributes to the persistent out-of-distribution (OOD) issues observed in existing methods; (2) an effective RoPE rescaling algorithm that adopts evolutionary search guided by \"needle-driven\" perplexity to address the insufficient training problem; (3) a mixed context window t"},"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":"2502.20082","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-27T13:41:07Z","cross_cats_sorted":[],"title_canon_sha256":"634f3e69d7d089c93048b8f1094446585f78a712689a9836846aeba9599ab40a","abstract_canon_sha256":"f84dd26b62988f895ff0431be21364e4da9a2a714ec6d48875c239aa40da4f92"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:21:09.276597Z","signature_b64":"JZ4HEG9/7hxI7mpduZ0GNNK4JLOiISYD6/NdlMLZWWxjZKkjT1cWqhVMkfdFsew8njMxAp+UwdC2pSrI5xxpCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5218e250c283f8995b9e0c1aefda5034881cd1e3425295972246ca27e732c495","last_reissued_at":"2026-07-05T10:21:09.276089Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:21:09.276089Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LongRoPE2: Near-Lossless LLM Context Window Scaling","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fan Yang, Gaokai Zhang, Gilsinia Lopez, Li Lyna Zhang, Mao Yang, Ning Shang, Siyuan Wang, Weizhu Chen","submitted_at":"2025-02-27T13:41:07Z","abstract_excerpt":"LongRoPE2 is a novel approach that extends the effective context window of pre-trained large language models (LLMs) to the target length, while preserving the performance on the original shorter context window. This is achieved by three contributions: (1) a hypothesis that insufficient training in higher RoPE dimensions contributes to the persistent out-of-distribution (OOD) issues observed in existing methods; (2) an effective RoPE rescaling algorithm that adopts evolutionary search guided by \"needle-driven\" perplexity to address the insufficient training problem; (3) a mixed context window t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.20082","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/2502.20082/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":"2502.20082","created_at":"2026-07-05T10:21:09.276149+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.20082v1","created_at":"2026-07-05T10:21:09.276149+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.20082","created_at":"2026-07-05T10:21:09.276149+00:00"},{"alias_kind":"pith_short_12","alias_value":"KIMOEUGCQP4J","created_at":"2026-07-05T10:21:09.276149+00:00"},{"alias_kind":"pith_short_16","alias_value":"KIMOEUGCQP4JSW46","created_at":"2026-07-05T10:21:09.276149+00:00"},{"alias_kind":"pith_short_8","alias_value":"KIMOEUGC","created_at":"2026-07-05T10:21:09.276149+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27705","citing_title":"Mitigating Position Bias in Transformers via Layer-Specific Positional Embedding Scaling","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2511.07328","citing_title":"Q-RAG: Long Context Multi-step Retrieval via Value-based Embedder Training","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20720","citing_title":"COMPASS: COntinual Multilingual PEFT with Adaptive Semantic Sampling","ref_index":60,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS","json":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS.json","graph_json":"https://pith.science/api/pith-number/KIMOEUGCQP4JSW46BQNO7WSQGS/graph.json","events_json":"https://pith.science/api/pith-number/KIMOEUGCQP4JSW46BQNO7WSQGS/events.json","paper":"https://pith.science/paper/KIMOEUGC"},"agent_actions":{"view_html":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS","download_json":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS.json","view_paper":"https://pith.science/paper/KIMOEUGC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.20082&json=true","fetch_graph":"https://pith.science/api/pith-number/KIMOEUGCQP4JSW46BQNO7WSQGS/graph.json","fetch_events":"https://pith.science/api/pith-number/KIMOEUGCQP4JSW46BQNO7WSQGS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS/action/storage_attestation","attest_author":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS/action/author_attestation","sign_citation":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS/action/citation_signature","submit_replication":"https://pith.science/pith/KIMOEUGCQP4JSW46BQNO7WSQGS/action/replication_record"}},"created_at":"2026-07-05T10:21:09.276149+00:00","updated_at":"2026-07-05T10:21:09.276149+00:00"}