{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BQ33AVXKQOY7W47ZLCIUI6ODU4","short_pith_number":"pith:BQ33AVXK","schema_version":"1.0","canonical_sha256":"0c37b056ea83b1fb73f958914479c3a738717859377c7d0a1be78d0fa417e024","source":{"kind":"arxiv","id":"2410.06722","version":3},"attestation_state":"computed","paper":{"title":"Scaling Laws For Mixed Quantization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Boyang Gu, Cheng Zhang, Jianqiao Lu, Jianyi Cheng, Pedro Gimenes, Xitong Gao, Yiren Zhao, Zeyu Cao","submitted_at":"2024-10-09T09:45:01Z","abstract_excerpt":"Post-training quantization of Large Language Models (LLMs) has proven effective in reducing the memory and computational requirements for inference. In this study, we focus on a straightforward question: When aiming for a target accuracy or perplexity with low-precision quantization, how much high-precision computation needs to be preserved, and how fine-grained this quantization would need to be as we scale LLMs to larger sizes? We first introduce two critical metrics, named the quantization ratio ($Q_r$) and quantization block size ($Q_b$). The former measures the number of parameters quanti"},"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":"2410.06722","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-09T09:45:01Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"9764f2db7e756141f25326cb279f44bf09515502ae11292d1b2788d8a3927fa1","abstract_canon_sha256":"450d72066b2c594b8a3a9424bf8cd1de77aeaa878ce26b2c9f8dc0799173c1ea"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:49:47.464578Z","signature_b64":"eGIMGU1IUbN4sPeavpGQPJqn5GJz7wcLxg/lW1uF2360ic68M2GJttBcF9aQlmoiTudAQHl8hOwSlkraa/zyDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0c37b056ea83b1fb73f958914479c3a738717859377c7d0a1be78d0fa417e024","last_reissued_at":"2026-07-05T11:49:47.464095Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:49:47.464095Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Laws For Mixed Quantization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Boyang Gu, Cheng Zhang, Jianqiao Lu, Jianyi Cheng, Pedro Gimenes, Xitong Gao, Yiren Zhao, Zeyu Cao","submitted_at":"2024-10-09T09:45:01Z","abstract_excerpt":"Post-training quantization of Large Language Models (LLMs) has proven effective in reducing the memory and computational requirements for inference. In this study, we focus on a straightforward question: When aiming for a target accuracy or perplexity with low-precision quantization, how much high-precision computation needs to be preserved, and how fine-grained this quantization would need to be as we scale LLMs to larger sizes? We first introduce two critical metrics, named the quantization ratio ($Q_r$) and quantization block size ($Q_b$). The former measures the number of parameters quanti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.06722","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/2410.06722/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":"2410.06722","created_at":"2026-07-05T11:49:47.464152+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.06722v3","created_at":"2026-07-05T11:49:47.464152+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.06722","created_at":"2026-07-05T11:49:47.464152+00:00"},{"alias_kind":"pith_short_12","alias_value":"BQ33AVXKQOY7","created_at":"2026-07-05T11:49:47.464152+00:00"},{"alias_kind":"pith_short_16","alias_value":"BQ33AVXKQOY7W47Z","created_at":"2026-07-05T11:49:47.464152+00:00"},{"alias_kind":"pith_short_8","alias_value":"BQ33AVXK","created_at":"2026-07-05T11:49:47.464152+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4","json":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4.json","graph_json":"https://pith.science/api/pith-number/BQ33AVXKQOY7W47ZLCIUI6ODU4/graph.json","events_json":"https://pith.science/api/pith-number/BQ33AVXKQOY7W47ZLCIUI6ODU4/events.json","paper":"https://pith.science/paper/BQ33AVXK"},"agent_actions":{"view_html":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4","download_json":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4.json","view_paper":"https://pith.science/paper/BQ33AVXK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.06722&json=true","fetch_graph":"https://pith.science/api/pith-number/BQ33AVXKQOY7W47ZLCIUI6ODU4/graph.json","fetch_events":"https://pith.science/api/pith-number/BQ33AVXKQOY7W47ZLCIUI6ODU4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4/action/storage_attestation","attest_author":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4/action/author_attestation","sign_citation":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4/action/citation_signature","submit_replication":"https://pith.science/pith/BQ33AVXKQOY7W47ZLCIUI6ODU4/action/replication_record"}},"created_at":"2026-07-05T11:49:47.464152+00:00","updated_at":"2026-07-05T11:49:47.464152+00:00"}