{"id":"arxiv/2606.29554","url":"https://arxiv.org/abs/2606.29554","tags":["cs.LG","math.NA","math.OC","stat.ML"],"score":0,"title":"Optimizer Memory Makes Shuffle Order a First-Order Source of Fine-Tuning Noise","domain":"cs.LG","source":"arxiv","authors":["John Sweeney"],"reviews":0,"subject":"cs.LG","summary":"Shuffle order can be a larger source of fine-tuning noise than a memoryless analysis predicts: fixed-clock optimizer memory makes local equal-multiset contrasts first order in the learning rate rather than second order, and the resulting order channel can be large enough for a single seed to flip a close A/B comparison. We isolate this mechanism and derive a fit-free way to size the noise it produces. For a memoryless optimizer, reordering an equal multiset has no first-order endpoint term; the leading local contrast is the $O(η^2)$ gradient bracket. Fixed-clock optimizers such as AdamW are…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-06-28","reactions":{}}