{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IYK7XYGVXBE73LWAFWLTLHNLTE","short_pith_number":"pith:IYK7XYGV","schema_version":"1.0","canonical_sha256":"4615fbe0d5b849fdaec02d97359dab993679ae839d35ad199f76f12c39cc883c","source":{"kind":"arxiv","id":"2410.08527","version":2},"attestation_state":"computed","paper":{"title":"Scaling Laws for Predicting Downstream Performance in LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Binxuan Huang, Heng Ji, Jingfeng Yang, Yangyi Chen, Yifan Gao, Zhengyang Wang","submitted_at":"2024-10-11T04:57:48Z","abstract_excerpt":"Precise estimation of downstream performance in large language models (LLMs) prior to training is essential for guiding their development process. Scaling laws analysis utilizes the statistics of a series of significantly smaller sampling language models (LMs) to predict the performance of the target LLM. For downstream performance prediction, the critical challenge lies in the emergent abilities in LLMs that occur beyond task-specific computational thresholds. In this work, we focus on the pre-training loss as a more computation-efficient metric for performance estimation. Our two-stage appro"},"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":"2410.08527","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-11T04:57:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"025b10b731ad6fbb4b985f88169d7817a7d5e60a3a91dfa27a5d2e0af5148bb6","abstract_canon_sha256":"f0700907d4f5be6a2507a6b727ed5fe7e9b09643d98a0acf1bf4530c4e4f6231"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:45:56.222342Z","signature_b64":"uQy5tFLZkUenmDEDt8af07N3GwVc5tw6YLoPM55htdwhAfm8IbztNePPHNzNJFuov4+P7ep3oHVc1mh3yAlPDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4615fbe0d5b849fdaec02d97359dab993679ae839d35ad199f76f12c39cc883c","last_reissued_at":"2026-07-05T10:45:56.221890Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:45:56.221890Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Laws for Predicting Downstream Performance in LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Binxuan Huang, Heng Ji, Jingfeng Yang, Yangyi Chen, Yifan Gao, Zhengyang Wang","submitted_at":"2024-10-11T04:57:48Z","abstract_excerpt":"Precise estimation of downstream performance in large language models (LLMs) prior to training is essential for guiding their development process. Scaling laws analysis utilizes the statistics of a series of significantly smaller sampling language models (LMs) to predict the performance of the target LLM. For downstream performance prediction, the critical challenge lies in the emergent abilities in LLMs that occur beyond task-specific computational thresholds. In this work, we focus on the pre-training loss as a more computation-efficient metric for performance estimation. Our two-stage appro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.08527","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/2410.08527/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":"2410.08527","created_at":"2026-07-05T10:45:56.221944+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.08527v2","created_at":"2026-07-05T10:45:56.221944+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.08527","created_at":"2026-07-05T10:45:56.221944+00:00"},{"alias_kind":"pith_short_12","alias_value":"IYK7XYGVXBE7","created_at":"2026-07-05T10:45:56.221944+00:00"},{"alias_kind":"pith_short_16","alias_value":"IYK7XYGVXBE73LWA","created_at":"2026-07-05T10:45:56.221944+00:00"},{"alias_kind":"pith_short_8","alias_value":"IYK7XYGV","created_at":"2026-07-05T10:45:56.221944+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.15079","citing_title":"Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2512.11470","citing_title":"Rethinking Expert Trajectory Utilization in LLM Post-training for Mathematical Reasoning","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2603.04816","citing_title":"Scaling Laws for Cross-Encoder Reranking","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14350","citing_title":"Distributionally Robust Multi-Task Reinforcement Learning via Adaptive Task Sampling","ref_index":245,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12705","citing_title":"Early Data Exposure Improves Robustness to Subsequent Fine-Tuning","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE","json":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE.json","graph_json":"https://pith.science/api/pith-number/IYK7XYGVXBE73LWAFWLTLHNLTE/graph.json","events_json":"https://pith.science/api/pith-number/IYK7XYGVXBE73LWAFWLTLHNLTE/events.json","paper":"https://pith.science/paper/IYK7XYGV"},"agent_actions":{"view_html":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE","download_json":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE.json","view_paper":"https://pith.science/paper/IYK7XYGV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.08527&json=true","fetch_graph":"https://pith.science/api/pith-number/IYK7XYGVXBE73LWAFWLTLHNLTE/graph.json","fetch_events":"https://pith.science/api/pith-number/IYK7XYGVXBE73LWAFWLTLHNLTE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE/action/storage_attestation","attest_author":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE/action/author_attestation","sign_citation":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE/action/citation_signature","submit_replication":"https://pith.science/pith/IYK7XYGVXBE73LWAFWLTLHNLTE/action/replication_record"}},"created_at":"2026-07-05T10:45:56.221944+00:00","updated_at":"2026-07-05T10:45:56.221944+00:00"}