Optimizer Memory Makes Shuffle Order a First-Order Source of Fine-Tuning Noise

• URL: http://arxiv.org/abs/2606.29554v1
• Date: Sun, 28 Jun 2026 18:44:08 GMT
• Title: Optimizer Memory Makes Shuffle Order a First-Order Source of Fine-Tuning Noise
• Abstract summary: Shuffle order can be a larger source of fine-tuning noise than a memoryless analysis predicts.<n>We isolate this mechanism and derive a fit-free way to size the noise it produces.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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 different. Their moment buffers, preconditioner state, and de-biasing counters advance with the step index rather than with the learning-rate-scaled time $τ=ηk$, so the same gradient can receive a position-dependent endpoint weight. For any fixed finite measurement window, a lifted-state expansion gives an $O(η)$ equal-multiset contrast whenever the first-order replay coefficient is nonzero, while regular and clock-matched controls remain $O(η^2)$; a bare fixed-$β$ momentum buffer is already enough. A bitwise-deterministic replay from one warmed optimizer state isolates the mechanism, giving order-variance slopes 1.83 for AdamW, 2.00 for fixed-$β$ momentum, and 4.00 for SGD; matching the memory clock to $τ$ restores the regular exponent. For AdamW with a frozen preconditioner, the same impulse-weight kernel gives a closed-form asymptotic order-variance floor after the local potentials are measured, with no fitted coefficients. The result is local to the measurement window (independent reshuffling can average the channel across windows), but it yields order-noise error bars, positional attribution weights, and a seed-budget criterion for fine-tuning comparisons.

Related papers

• Decay versus dephasing in Rydberg analog optimization: exchange rate, mechanism, and schedule design [0.05427878000393745]
  Noiseless Rydberg-atom optimization is studied in terms of decoherence and dephasing.<n>First-order perturbation theory reproduces $$ from noiseless propagation alone.<n>Because the whole cost model is noiseless, a schedule can be tuned to a device's channel mixture without any noisy simulation.
  arXiv  Detail & Related papers  (2026-08-30T15:41:02Z)
• Second-Moment Memory in Coordinatewise Adam [0.0]
  We show that second-moment memory can itself suppress progress toward optimum gradients.<n>Long second-moment memory can slow optimization even when the gradient noise has finite variance.
  arXiv  Detail & Related papers  (2026-08-16T15:56:36Z)
• Asymptotic-Preserving A Posteriori Analysis of Diffusion and Flow-Matching Samplers [0.6768558752130311]
  Diffusion and flow-matching samplers integrate a learned probability-flow ODE from a large noise scale down to a small terminal floor $_min$, at which the score is stiff and the flow develops a boundary layer.<n>We treat $_min$ as a singular-perturbation parameter and determine which fixed-step samplers are solvable.
  arXiv  Detail & Related papers  (2026-07-05T04:48:33Z)
• Robust Layerwise Scaling Rules by Proper Weight Decay Tuning [50.11170157029911]
  In modern scale-invariant architectures, training quickly enters an degrading-governed steady state.<n>We introduce a weight-decay scaling rule for AdamW that preserves sublayer gain across widths.<n>Our results extend $mu$P beyond the near-init regime by explicitly controlling the steady-state scales set by parameters.
  arXiv  Detail & Related papers  (2025-10-17T02:58:35Z)
• Proving the Limited Scalability of Centralized Distributed Optimization via a New Lower Bound Construction [57.93371273485736]
  We consider a centralized distributed learning setup where all workers jointly find an unbiased bound LDeltaepsilon2,$ better poly-logarithmically in $n$, even in the homogeneous (i.i.d.) case, where all workers access the same distribution.
  arXiv  Detail & Related papers  (2025-06-30T13:27:39Z)
• VAMO: Efficient Zeroth-Order Variance Reduction for SGD with Faster Convergence [6.574641780732972]
  Large-scale non problems are common in deep learning.<n>First-order (FO)s serve as today's baselines.<n>ZO algorithms reduce the size of computation and memory costs.<n>VAMO achieves these gains with smaller dynamic memory requirements.
  arXiv  Detail & Related papers  (2025-05-20T05:31:15Z)
• Beyond likelihood ratio bias: Nested multi-time-scale stochastic approximation for likelihood-free parameter estimation [49.78792404811239]
  We study inference in simulation-based models where the analytical form of the likelihood is unknown.<n>We use a ratio-free nested multi-time-scale approximation (SA) method that simultaneously tracks the score and drives the parameter update.<n>We show that our algorithm can eliminate the original bias $Obig(sqrtfrac1Nbig)$ and accelerate the convergence rate from $Obig(beta_k+sqrtfracalpha_kNbig)$.
  arXiv  Detail & Related papers  (2024-11-20T02:46:15Z)
• Semi-Discrete Optimal Transport: Nearly Minimax Estimation With Stochastic Gradient Descent and Adaptive Entropic Regularization [38.67914746910537]
  We prove an $mathcalO(t-1)$ lower bound rate for the OT map, using the similarity between Laguerre cells estimation and density support estimation.<n>To nearly achieve the desired fast rate, we design an entropic regularization scheme decreasing with the number of samples.
  arXiv  Detail & Related papers  (2024-05-23T11:46:03Z)
• Extra-Newton: A First Approach to Noise-Adaptive Accelerated Second-Order Methods [57.050204432302195]
  This work proposes a universal and adaptive second-order method for minimizing second-order smooth, convex functions. Our algorithm achieves $O(sigma / sqrtT)$ convergence when the oracle feedback is with variance $sigma2$, and improves its convergence to $O( 1 / T3)$ with deterministic oracles.
  arXiv  Detail & Related papers  (2022-11-03T14:12:51Z)
• Sharper Convergence Guarantees for Asynchronous SGD for Distributed and Federated Learning [77.22019100456595]
  We show a training algorithm for distributed computation workers with varying communication frequency. In this work, we obtain a tighter convergence rate of $mathcalO!!!(sigma2-2_avg!! . We also show that the heterogeneity term in rate is affected by the average delay within each worker.
  arXiv  Detail & Related papers  (2022-06-16T17:10:57Z)
• Asynchronous Stochastic Optimization Robust to Arbitrary Delays [54.61797739710608]
  We consider optimization with delayed gradients where, at each time stept$, the algorithm makes an update using a stale computation - d_t$ for arbitrary delay $d_t gradient. Our experiments demonstrate the efficacy and robustness of our algorithm in cases where the delay distribution is skewed or heavy-tailed.
  arXiv  Detail & Related papers  (2021-06-22T15:50:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.