cs.LGNov 2, 2026

A Scaling Theory of Multi-Epoch Pretraining: Late-Time Overfitting and Ensembling

Authors: Zhen Yang, Yidi Miao, Blake Bordelon

Organizations: University of Texas at Austin · Carnegie Mellon University

Abstract

In the near future language model pretraining will become bound by total data rather than total compute. Such a regime requires revisiting scaling laws and scaling law theory. Motivated by prior works that have found utility in ensembling, we study how the number of training epochs and ensemble size jointly shape generalization when training data are fixed. In language-model experiments on FineWeb, logit averaging reduces late-training deterioration, delays optimal stopping, and can outperform wider or deeper single models at matched parameter-token budgets over a common training horizon. To analyze this interaction between data reuse and ensembling, we derive a dynamical mean-field theory (DMFT) for multi-epoch stochastic gradient descent in power-law random-feature ensembles, tracking error correlations across epochs and models. Ensembling reduces the prediction variance responsible for late-time growth in expected test error, changing the optimal stopping time. We use these stopping times to derive compute-allocation laws at fixed data and individual model sizes. Under this allocation, ensemble size grows with compute at least as fast as training time per member, while excess test loss above the limiting loss decreases as a power of compute.

Figures & tables

Appendix figures & tables18 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Jun 2, 2026

q0: Primitives for Hyper-Epoch Pretraining

Multi-epoch training is becoming the standard now that compute is growing faster than the supply of high-quality text. But pretraining a single model saturates within a few passes, long before the compute budget is exhausted. We argue this calls for a conceptual shift from training a single model toward exploring a population of models and aggregating their predictions. We introduce hyper-epoch pretraining (q0), which turns a multi-epoch budget into a population of diverse models whose combined predictions reach a lower validation loss than a single refined model. q0 reduces to three core primitives. A cyclic schedule with anti-correlated learning rate and weight decay collects diverse models from a few parallel trajectories. Chain distillation trains each model against its predecessor so that model quality compounds across the population. A learned prior, fit on a held out set, selects and weights members for any inference budget. On a 1.8B-parameter model trained on 100M FineWeb tokens, q0 matches a strong 256-epoch ensemble baseline using only ~56 epochs (~4.6x fewer), or ~67 epochs (~3.8x fewer) when matched to the baseline's ensemble size, and continues to improve beyond it. These gains reach cumulative ~12.9x data efficiency under the Slowrun setting and transfer to downstream benchmarks. Crucially, the optimal allocation shifts with the budget, so we give prescriptive recipes for how to spend a given epoch budget to maximize generalization, from a single epoch up to the largest budgets.
Oct 27, 2026

Improving Data-Constrained Scaling with Hyperparameter Transfer and Ensembling

As growing compute budgets outpace the growth of high-quality training data, large language models must be trained with both compute and data constraints in mind. Unfortunately, existing prescriptions for data-constrained scaling suffer from two important limitations. First, past works hold key hyperparameters such as weight decay fixed with scale, despite the fact that optimal hyperparameters vary wildly with model size and number of epochs. Second, past works neglect ensembling, a promising axis for scaling under data constraints. As a surprising consequence, prior prescriptions can lead to loss that increases with compute budget. We address both limitations by fitting power laws that prescribe how to transfer hyperparameters with scale and scaling laws that prescribe how to allocate compute with ensembling. Training on 100M tokens from FineWeb, our recipe reduces validation loss by over 0.15 nats compared to prior recipes, a 14% decrease in perplexity, and achieves monotonic improvements with compute through ensembling. These gains are achieved using prescriptions fit to smaller runs and validated at up to 12×\times larger compute budgets.
May 9, 2026

Practical Scaling Laws: Converting Compute into Performance in a Data-Constrained World

The scaling laws guiding modern model training were calibrated for a single regime: data-rich, single-epoch pretraining. The dominant such scaling law form, Chinchilla's L=E+A/Nα+B/DβL = E + A/N^α+ B/D^β, has three structural limitations outside that regime: it diverges as unique data shrinks instead of saturating at the uninformed baseline; it cannot represent overfitting when capacity exceeds the data; and it conflates total examples seen with unique examples available. We propose a closed-form extension, L(N,D,T)=E+(L0−E) h/(1+h)L(N, D, T) = E + (L_0 - E)\,h/(1+h) with h=a/Nα+b/Tβ+c Nγ/Dδh = a/N^α+ b/T^β+ c\,N^γ/D^δ, that decomposes loss into undercapacity, undertraining, and overfitting terms. It saturates between the irreducible loss EE and an uninformed baseline L0L_0 fixed by the loss type, and reduces to Chinchilla in the data-rich, single-epoch limit. We validate it on four multi-epoch experiments spanning four architecture families (MLPs, ResNets, Fourier neural operators, and transformers) across vision, scientific ML, and language domains, and refit it to five published LLM scaling-law grids. Extrapolating to higher compute and larger unique data than seen at fit time, our form achieves state-of-the-art RMSE on every published LLM grid we evaluate and on most cells of our constructed experiments. Once calibrated, the form admits a cost-aware allocation that recovers Chinchilla's optimum when data is free and shifts toward smaller corpora and more epochs as data grows expensive.