{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NDSOJ2IGN5IOBZJYU4FEVXR6EE","short_pith_number":"pith:NDSOJ2IG","schema_version":"1.0","canonical_sha256":"68e4e4e9066f50e0e538a70a4ade3e2104bc5ef0dba77960c9b59ad0a69d10b9","source":{"kind":"arxiv","id":"2605.26895","version":1},"attestation_state":"computed","paper":{"title":"Negligible in Size, Significant in Effect: On Scale Vectors in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Binghui Li, Kai Shen, Mingze Wang, Shuchen Zhu, Shu Zhong, Yuxin Fang","submitted_at":"2026-05-26T11:56:49Z","abstract_excerpt":"Normalization layers in modern large language models (LLMs) consist of a deterministic normalization operation and a learnable scale vector. While the normalization operation has been extensively studied, the scale vector remains poorly understood despite its ubiquitous use. In this work, we present a systematic study of scale vectors in LLMs from the perspectives of expressivity, optimization, and architectural structure. First, we show empirically that although scale vectors constitute only a negligible fraction of model parameters, removing them substantially degrades LLM pre-training. Our "},"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":"2605.26895","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-26T11:56:49Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"33786f2f44f56050a2fb43006b97b21b062172ab71b5214517c64e1c5abdddc2","abstract_canon_sha256":"4cdc8025e8a8110a7c67da14aaac4cde555740671e756d7846b0efc73c1b49b5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:06:18.521827Z","signature_b64":"7BhVnz2HQtyL7Edy3kC4yVlTVnTg95HdRg1+XABawhtqZGUfVOdnfdbxOrwD5RF7i+//HNNoaZWGr6sYRRe1Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"68e4e4e9066f50e0e538a70a4ade3e2104bc5ef0dba77960c9b59ad0a69d10b9","last_reissued_at":"2026-05-27T01:06:18.521112Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:06:18.521112Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Negligible in Size, Significant in Effect: On Scale Vectors in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Binghui Li, Kai Shen, Mingze Wang, Shuchen Zhu, Shu Zhong, Yuxin Fang","submitted_at":"2026-05-26T11:56:49Z","abstract_excerpt":"Normalization layers in modern large language models (LLMs) consist of a deterministic normalization operation and a learnable scale vector. While the normalization operation has been extensively studied, the scale vector remains poorly understood despite its ubiquitous use. In this work, we present a systematic study of scale vectors in LLMs from the perspectives of expressivity, optimization, and architectural structure. First, we show empirically that although scale vectors constitute only a negligible fraction of model parameters, removing them substantially degrades LLM pre-training. Our "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26895","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/2605.26895/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":"2605.26895","created_at":"2026-05-27T01:06:18.521220+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.26895v1","created_at":"2026-05-27T01:06:18.521220+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26895","created_at":"2026-05-27T01:06:18.521220+00:00"},{"alias_kind":"pith_short_12","alias_value":"NDSOJ2IGN5IO","created_at":"2026-05-27T01:06:18.521220+00:00"},{"alias_kind":"pith_short_16","alias_value":"NDSOJ2IGN5IOBZJY","created_at":"2026-05-27T01:06:18.521220+00:00"},{"alias_kind":"pith_short_8","alias_value":"NDSOJ2IG","created_at":"2026-05-27T01:06:18.521220+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/NDSOJ2IGN5IOBZJYU4FEVXR6EE","json":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE.json","graph_json":"https://pith.science/api/pith-number/NDSOJ2IGN5IOBZJYU4FEVXR6EE/graph.json","events_json":"https://pith.science/api/pith-number/NDSOJ2IGN5IOBZJYU4FEVXR6EE/events.json","paper":"https://pith.science/paper/NDSOJ2IG"},"agent_actions":{"view_html":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE","download_json":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE.json","view_paper":"https://pith.science/paper/NDSOJ2IG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.26895&json=true","fetch_graph":"https://pith.science/api/pith-number/NDSOJ2IGN5IOBZJYU4FEVXR6EE/graph.json","fetch_events":"https://pith.science/api/pith-number/NDSOJ2IGN5IOBZJYU4FEVXR6EE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE/action/storage_attestation","attest_author":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE/action/author_attestation","sign_citation":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE/action/citation_signature","submit_replication":"https://pith.science/pith/NDSOJ2IGN5IOBZJYU4FEVXR6EE/action/replication_record"}},"created_at":"2026-05-27T01:06:18.521220+00:00","updated_at":"2026-05-27T01:06:18.521220+00:00"}