{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:CAI4IMOSTWORB26XEK3HMLNKKA","short_pith_number":"pith:CAI4IMOS","schema_version":"1.0","canonical_sha256":"1011c431d29d9d10ebd722b6762daa50049629f32c1a69c94a3fd0ed1f29b933","source":{"kind":"arxiv","id":"2312.03732","version":1},"attestation_state":"computed","paper":{"title":"A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LoRA adapters should be scaled by dividing by the square root of the rank rather than the full rank to stabilize learning.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Damjan Kalajdzievski","submitted_at":"2023-11-28T03:23:20Z","abstract_excerpt":"As large language models (LLMs) have become increasingly compute and memory intensive, parameter-efficient fine-tuning (PEFT) methods are now a common strategy to fine-tune LLMs. A popular PEFT method is Low-Rank Adapters (LoRA), which adds trainable low-rank \"adapters\" to selected layers. Each adapter consists of a low-rank matrix product, multiplicatively scaled by a rank-dependent factor. This scaling factor, which divides adapters by a factor of the rank, results in slowed learning and stunted performance for LoRA with higher-rank adapters. Consequently, the use of LoRA in practice has gen"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2312.03732","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-28T03:23:20Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"af5df5ac73aa5e580bcc05a69a919e9f79e524916c2b8436161eafa0c94e374d","abstract_canon_sha256":"fc96d70f4339ef6a7ea4f91c0a84441b8dc996a69d85a3dd76178a6f181d107c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:50.003179Z","signature_b64":"UqMXcBRQltHaEhdzJWNF3thAJXl7P1sLwmO2B5ZY3z8fxiwyCnct463CWYxDFYTOtNSc43XnX/w1qlUnPApWDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1011c431d29d9d10ebd722b6762daa50049629f32c1a69c94a3fd0ed1f29b933","last_reissued_at":"2026-05-17T23:38:50.002668Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:50.002668Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LoRA adapters should be scaled by dividing by the square root of the rank rather than the full rank to stabilize learning.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Damjan Kalajdzievski","submitted_at":"2023-11-28T03:23:20Z","abstract_excerpt":"As large language models (LLMs) have become increasingly compute and memory intensive, parameter-efficient fine-tuning (PEFT) methods are now a common strategy to fine-tune LLMs. A popular PEFT method is Low-Rank Adapters (LoRA), which adds trainable low-rank \"adapters\" to selected layers. Each adapter consists of a low-rank matrix product, multiplicatively scaled by a rank-dependent factor. This scaling factor, which divides adapters by a factor of the rank, results in slowed learning and stunted performance for LoRA with higher-rank adapters. Consequently, the use of LoRA in practice has gen"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we study the impact of the scaling factor on the learning process and prove that LoRA adapters should be divided by a factor of the square root of the rank","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The proof that 1/sqrt(rank) is optimal rests on unstated assumptions about initialization variance, gradient flow, and the precise form of the LoRA update rule during fine-tuning; these assumptions are not detailed in the provided abstract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LoRA adapters should be scaled by 1/sqrt(rank) rather than 1/rank to stabilize learning and enable effective use of higher ranks during fine-tuning of large language models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LoRA adapters should be scaled by dividing by the square root of the rank rather than the full rank to stabilize learning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"228ad1f7685270356943d91311780a58336dccb02bf917c3b10a87eac8f2f96d"},"source":{"id":"2312.03732","kind":"arxiv","version":1},"verdict":{"id":"c9b1eb0e-c4c8-43ef-a1df-9e713c8a8353","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T22:02:04.412118Z","strongest_claim":"we study the impact of the scaling factor on the learning process and prove that LoRA adapters should be divided by a factor of the square root of the rank","one_line_summary":"LoRA adapters should be scaled by 1/sqrt(rank) rather than 1/rank to stabilize learning and enable effective use of higher ranks during fine-tuning of large language models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The proof that 1/sqrt(rank) is optimal rests on unstated assumptions about initialization variance, gradient flow, and the precise form of the LoRA update rule during fine-tuning; these assumptions are not detailed in the provided abstract.","pith_extraction_headline":"LoRA adapters should be scaled by dividing by the square root of the rank rather than the full rank to stabilize learning."},"references":{"count":91,"sample":[{"doi":"","year":2023,"title":"2023 , publisher =","work_id":"4e8b0b95-ae6c-4edf-8218-6b1fbbfd9734","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.5281/zenodo.5371628","year":null,"title":"Aaron Gokaslan, Vanya Cohen, Ellie Pavlick, and Stefanie Tellex","work_id":"27d00efc-1c8b-4af4-bd30-5dfa575d0989","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge , author=. 2018 , eprint=","work_id":"9880c11f-6c71-4e36-9736-affdf3d20344","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"HellaSwag: Can a Machine Really Finish Your Sentence? , author=. 2019 , eprint=","work_id":"b5234106-7b98-47af-87c3-45d428066901","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Measuring Massive Multitask Language Understanding , author=. 2021 , eprint=","work_id":"01eb26fa-b178-499b-bc7b-b2ff5dd67b24","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":91,"snapshot_sha256":"fea4b5c0f7a8f09e836708a7210ddf6567efda57631d4dc9edc69c6adae48df3","internal_anchors":6},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1fded5ab56adef58140b9f279197ddf9fd002341cffcd2bd1c2d972b2a474ac4"},"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":"2312.03732","created_at":"2026-05-17T23:38:50.002749+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.03732v1","created_at":"2026-05-17T23:38:50.002749+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.03732","created_at":"2026-05-17T23:38:50.002749+00:00"},{"alias_kind":"pith_short_12","alias_value":"CAI4IMOSTWOR","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_16","alias_value":"CAI4IMOSTWORB26X","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_8","alias_value":"CAI4IMOS","created_at":"2026-05-18T12:33:33.725879+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":24,"internal_anchor_count":24,"sample":[{"citing_arxiv_id":"2605.07111","citing_title":"Beyond LoRA vs. Full Fine-Tuning: Gradient-Guided Optimizer Routing for LLM Adaptation","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16470","citing_title":"Strategic Over-Parameterization for Generalizable Low-Rank Adaptation","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15916","citing_title":"LoCO: Low-rank Compositional Rotation Fine-tuning","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2506.01897","citing_title":"MLorc: Momentum Low-rank Compression for Memory Efficient Large Language Model Adaptation","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2603.21011","citing_title":"ALL-FEM: Agentic Large Language models Fine-tuned for Finite Element Methods","ref_index":70,"is_internal_anchor":true},{"citing_arxiv_id":"2602.23024","citing_title":"InCoM: Intent-Driven Perception and Structured Coordination for Mobile Manipulation","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2603.15045","citing_title":"LLMs and Speech: Integration vs. Combination","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14217","citing_title":"PreFT: Prefill-only finetuning for efficient inference","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2604.14171","citing_title":"Benchmarking Linguistic Adaptation in Comparable-Sized LLMs: A Study of Llama-3.1-8B, Mistral-7B-v0.1, and Qwen3-8B on Romanized Nepali","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12207","citing_title":"Not How Many, But Which: Parameter Placement in Low-Rank Adaptation","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26388","citing_title":"SplitFT: An Adaptive Federated Split Learning System For LLMs Fine-Tuning","ref_index":52,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26136","citing_title":"One Voice, Many Tongues: Cross-Lingual Voice Cloning for Scientific Speech","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10468","citing_title":"Can Muon Fine-tune Adam-Pretrained Models?","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09015","citing_title":"LLiMba: Sardinian on a Single GPU -- Adapting a 3B Language Model to a Vanishing Romance Language","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.23665","citing_title":"HAC: Parameter-Efficient Hyperbolic Adaptation of CLIP for Zero-Shot VQA","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06175","citing_title":"VLA-GSE: Boosting Parameter-Efficient Fine-Tuning in VLA with Generalized and Specialized Experts","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"2604.08749","citing_title":"A Little Rank Goes a Long Way: Random Scaffolds with LoRA Adapters Are All You Need","ref_index":34,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06175","citing_title":"VLA-GSE: Boosting Parameter-Efficient Fine-Tuning in VLA with Generalized and Specialized Experts","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07111","citing_title":"Beyond LoRA vs. Full Fine-Tuning: Gradient-Guided Optimizer Routing for LLM Adaptation","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07850","citing_title":"MatryoshkaLoRA: Learning Accurate Hierarchical Low-Rank Representations for LLM Fine-Tuning","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2604.14356","citing_title":"When PCOS Meets Eating Disorders: An Explainable AI Approach to Detecting the Hidden Triple Burden","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2604.18124","citing_title":"TLoRA: Task-aware Low Rank Adaptation of Large Language Models","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2604.17051","citing_title":"Efficient Task Adaptation in Large Language Models via Selective Parameter Optimization","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2604.17961","citing_title":"DifFoundMAD: Foundation Models meet Differential Morphing Attack Detection","ref_index":28,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA","json":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA.json","graph_json":"https://pith.science/api/pith-number/CAI4IMOSTWORB26XEK3HMLNKKA/graph.json","events_json":"https://pith.science/api/pith-number/CAI4IMOSTWORB26XEK3HMLNKKA/events.json","paper":"https://pith.science/paper/CAI4IMOS"},"agent_actions":{"view_html":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA","download_json":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA.json","view_paper":"https://pith.science/paper/CAI4IMOS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.03732&json=true","fetch_graph":"https://pith.science/api/pith-number/CAI4IMOSTWORB26XEK3HMLNKKA/graph.json","fetch_events":"https://pith.science/api/pith-number/CAI4IMOSTWORB26XEK3HMLNKKA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA/action/storage_attestation","attest_author":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA/action/author_attestation","sign_citation":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA/action/citation_signature","submit_replication":"https://pith.science/pith/CAI4IMOSTWORB26XEK3HMLNKKA/action/replication_record"}},"created_at":"2026-05-17T23:38:50.002749+00:00","updated_at":"2026-05-17T23:38:50.002749+00:00"}