q0: Primitives for Hyper-Epoch Pretraining

• URL: http://arxiv.org/abs/2606.03938v2
• Date: Wed, 03 Jun 2026 02:07:12 GMT
• Title: q0: Primitives for Hyper-Epoch Pretraining
• Abstract summary: Pretraining a single model saturates within a few passes, long before the compute budget is exhausted.<n>We introduce hyper-epoch pretraining (q0), which turns a multi-epoch budget into a population of diverse models.<n>We show that 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.
• Score: 0.5980755233352995
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• Unextractable Protocol Models: Collaborative Training and Inference without Weight Materialization [58.14514930760722]
  We consider a decentralized setup in which the participants collaboratively train and serve a large neural network.<n>In this setup, we explore the possibility of unmaterializable weights, where a full weight set is never available to any one participant.<n>We introduce Unextractable Protocol Models (UPMs): a training and inference framework that leverages the sharded model setup.
  arXiv  Detail & Related papers  (2026-05-22T10:24:57Z)
• IMU-1: Sample-Efficient Pre-training of Small Language Models [0.0]
  We present IMU-1, a 430M- parameter language model trained on 72B tokens that approaches the benchmark performance of models trained on 56x more data.<n>We describe a validated training recipe combining recent architectural interventions (QK-norm attention, per-head gating, value residuals, LayerNorm scaling) with optimization advances.
  arXiv  Detail & Related papers  (2026-01-25T21:24:15Z)
• Elastic ViTs from Pretrained Models without Retraining [74.5386166956142]
  Vision foundation models achieve remarkable performance but are only available in a limited set of pre-determined sizes.<n>We introduce SnapViT: Single-shot network approximation for pruned Vision Transformers.<n>Our approach efficiently combines gradient information with cross-network structure correlations, approximated via an evolutionary algorithm.
  arXiv  Detail & Related papers  (2025-10-20T16:15:03Z)
• Catch Your Breath: Adaptive Computation for Self-Paced Sequence Production [55.76222360698305]
  We explore a class of supervised training objectives that allow a language model to dynamically and autonomously scale the number of compute steps used for each input token.<n>For any token, the model can request additional compute steps by emitting a don't know> output.<n>We find that the CYB model requests additional steps when doing so improves accuracy, and the model adapts its processing time to token-level complexity and context.
  arXiv  Detail & Related papers  (2025-10-13T21:07:05Z)
• CAST: Continuous and Differentiable Semi-Structured Sparsity-Aware Training for Large Language Models [27.682531424487564]
  Sparsity-aware training is an effective approach for transforming large language models into hardware-friendly sparse patterns.<n>We propose Continuous Adaptive Sparse Trainer (CAST), a continuous and differentiable sparsity-aware training framework for sparse models.<n>Our results demonstrate significant improvements over previous state-of-the-art methods in both perplexity and zero-shot accuracy with minimal training resources.
  arXiv  Detail & Related papers  (2025-09-30T09:28:47Z)
• Pre-training under infinite compute [87.02472603429936]
  We show that data-constrained approaches of increasing epoch count and parameter count eventually overfit.<n>We identify that ensembling independently trained models achieves a significantly lower loss asymptote than the regularized recipe.<n>Our results show that simple algorithmic improvements can enable significantly more data-efficient pre-training in a compute-rich future.
  arXiv  Detail & Related papers  (2025-09-18T09:36:23Z)
• Intention-Conditioned Flow Occupancy Models [80.42634994902858]
  Large-scale pre-training has fundamentally changed how machine learning research is done today.<n>Applying this same framework to reinforcement learning is appealing because it offers compelling avenues for addressing core challenges in RL.<n>Recent advances in generative AI have provided new tools for modeling highly complex distributions.
  arXiv  Detail & Related papers  (2025-06-10T15:27:46Z)
• Zeroth-Order Adaptive Neuron Alignment Based Pruning without Re-Training [7.972074133591484]
  We propose textbfNeuroAl, a emphtop-up algorithm that can be used on top of any given pruning algorithm for LLMs.<n>Our method consistently outperforms the latest state-of-the-art methods in terms of performance-runtime trade-off.
  arXiv  Detail & Related papers  (2024-11-11T15:30:16Z)
• Efficient Training of Language Models with Compact and Consistent Next Token Distributions [23.312920633391837]
  We show that we can train better models faster by pre-aggregating the corpus with a collapsed $n$-gram distribution. Our approximation facilitates scalability of gains to larger datasets and models.
  arXiv  Detail & Related papers  (2024-07-03T05:40:41Z)
• Language models scale reliably with over-training and on downstream tasks [121.69867718185125]
  Scaling laws are useful guides for derisking expensive training runs. However, there remain gaps between current studies and how language models are trained. In contrast, scaling laws mostly predict loss on inference, but models are usually compared on downstream task performance.
  arXiv  Detail & Related papers  (2024-03-13T13:54:00Z)
• RanPAC: Random Projections and Pre-trained Models for Continual Learning [59.07316955610658]
  Continual learning (CL) aims to learn different tasks (such as classification) in a non-stationary data stream without forgetting old ones. We propose a concise and effective approach for CL with pre-trained models.
  arXiv  Detail & Related papers  (2023-07-05T12:49:02Z)
• Finding the SWEET Spot: Analysis and Improvement of Adaptive Inference in Low Resource Settings [6.463202903076821]
  We compare the two main approaches for adaptive inference, Early-Exit and Multi-Model, when training data is limited. Early-Exit provides a better speed-accuracy trade-off due to the overhead of the Multi-Model approach. We propose SWEET, an Early-Exit fine-tuning method that assigns each classifier its own set of unique model weights.
  arXiv  Detail & Related papers  (2023-06-04T09:16:39Z)
• Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints [59.39280540478479]
  We propose sparse upcycling -- a simple way to reuse sunk training costs by initializing a sparsely activated Mixture-of-Experts model from a dense checkpoint. We show that sparsely upcycled T5 Base, Large, and XL language models and Vision Transformer Base and Large models, respectively, significantly outperform their dense counterparts on SuperGLUE and ImageNet.
  arXiv  Detail & Related papers  (2022-12-09T18:57:37Z)
• Not All Models Are Equal: Predicting Model Transferability in a Self-challenging Fisher Space [51.62131362670815]
  This paper addresses the problem of ranking the pre-trained deep neural networks and screening the most transferable ones for downstream tasks. It proposes a new transferability metric called textbfSelf-challenging textbfFisher textbfDiscriminant textbfAnalysis (textbfSFDA)
  arXiv  Detail & Related papers  (2022-07-07T01:33:25Z)
• Deep Ensembles for Low-Data Transfer Learning [21.578470914935938]
  We study different ways of creating ensembles from pre-trained models. We show that the nature of pre-training itself is a performant source of diversity. We propose a practical algorithm that efficiently identifies a subset of pre-trained models for any downstream dataset.
  arXiv  Detail & Related papers  (2020-10-14T07:59:00Z)

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.