In ImageNet, our method achieves speedups of 4 in MoCo, 3.3 in SimCLR, and 2.5 in DINO, demonstrating substantial efficiency gains.

We instead propose to estimate the final test performance based on a simple measure of training speed.

The expanding computational costs and limited resources underscore the critical need for budgeted-iteration training, which aims to achieve optimal learning within predetermined iteration budgets. While learning rate schedules fundamentally govern the performance of different networks and tasks, particularly in budgeted-iteration scenarios, their design remains largely heuristic, lacking theoretical foundations. In addition, the optimal learning rate schedule requires extensive trial-and-error selection, making the training process inefficient. In this work, we propose the Unified Budget-Aware (UBA) schedule, a theoretically grounded learning rate schedule that consistently outperforms commonly-used schedules among diverse architectures and tasks under different constrained training budgets. First, we bridge the gap by constructing a novel training budget-aware optimization framework, which explicitly accounts for the robustness to landscape curvature variations. From this framework, we derive the UBA schedule, controlled by a single hyper-parameter φthat provides a trade-off between flexibility and simplicity, eliminating the need for per-network numerical optimization. Moreover, we establish a theoretical connection between φand the condition number, adding interpretation and justification to our approach. Besides, we prove the convergence for different values of φ. We offer practical guidelines for its selection via theoretical analysis and empirical results. Extensive experimental results show that UBA consistently surpasses the commonly-used schedules across diverse vision and language tasks, spanning network architectures (e.g., ResNet, OLMo) and scales, under different training-iteration budgets.

Consistency Models (CMs) have shown promise for efficient one-step generation. However, most existing CMs rely on manually designed discretization schemes, which can cause repeated adjustments for different noise schedules and datasets. To address this, we propose a unified framework for the automatic and adaptive discretization of CMs, formulating it as an optimization problem with respect to the discretization step. Concretely, during the consistency training process, we propose using local consistency as the optimization objective to ensure trainability by avoiding excessive discretization, and taking global consistency as a constraint to ensure stability by controlling the denoising error in the training target. We establish the trade-off between local and global consistency with a Lagrange multiplier. Building on this framework, we achieve adaptive discretization for CMs using the Gauss-Newton method. We refer to our approach as ADCMs. Experiments demonstrate that ADCMs significantly improve the training efficiency of CMs, achieving superior generative performance with minimal training overhead on both CIFAR-10 and ImageNet. Moreover, ADCMs exhibit strong adaptability to more advanced DM variants. Code is available at https://github.com/rainstonee/ADCM.

In ImageNet, our method achieves speedups of 4 in MoCo, 3.3 in SimCLR, and 2.5 in DINO, demonstrating substantial efficiency gains.

To all reviewers, thank you very much for your thoughtful comments and suggestions. R#1: "...importance of similarity among the selected tasks... " R#1: "...domain randomization, when enough samples are used, is a better alternative to meta-learning... " R#2: "...Theorems 1 and 2 are asymptotic... " Hence, the theorems are NOT asymptotic. We will remove the asymptotic parts for clarity. R#2: 'Assumption 2 ... the per-task optimal models are centered around the corresponding optimal solutions. This assumption can easily be dropped with the cost of including the distance as a term.

Reliable yet efficient evaluation of generalisation performance of a proposed architecture is crucial to the success of neural architecture search (NAS).

The past decade has seen incredible scaling of AI systems by a few companies, leading to inequality in AI model performance. This paper argues that, contrary to prevailing intuition, the diminishing returns to compute scaling will lead to a convergence of AI model capabilities. In other words, meek models (those with limited computation budget) shall inherit the earth, approaching the performance level of the best models overall. We develop a model illustrating that under a fixed-distribution next-token objective, the marginal capability returns to raw compute shrink substantially. Given current scaling practices, we argue that these diminishing returns are strong enough that even companies that can scale their models exponentially faster than other organizations will eventually have little advantage in capabilities. As part of our argument, we give several reasons that proxies like training loss differences capture important capability measures using evidence from benchmark data and theoretical performance models. In addition, we analyze empirical data on the capability difference of AI models over time. Finally, in light of the increasing ability of meek models, we argue that AI strategy and policy require reexamination, and we outline the areas this shift will affect.

Selecting high-quality pre-training data for large language models (LLMs) is crucial for enhancing their overall performance under limited computation budget, improving both training and sample efficiency. Recent advancements in file selection primarily rely on using an existing or trained proxy model to assess the similarity of samples to a target domain, such as high quality sources BookCorpus and Wikipedia. However, upon revisiting these methods, the domain-similarity selection criteria demonstrates a diversity dilemma, i.e.dimensional collapse in the feature space, improving performance on the domain-related tasks but causing severe degradation on generic performance. To prevent collapse and enhance diversity, we propose a DiverSified File selection algorithm (DiSF), which selects the most decorrelated text files in the feature space. We approach this with a classical greedy algorithm to achieve more uniform eigenvalues in the feature covariance matrix of the selected texts, analyzing its approximation to the optimal solution under a formulation of $γ$-weakly submodular optimization problem. Empirically, we establish a benchmark and conduct extensive experiments on the TinyLlama architecture with models from 120M to 1.1B parameters. Evaluating across nine tasks from the Harness framework, DiSF demonstrates a significant improvement on overall performance. Specifically, DiSF saves 98.5% of 590M training files in SlimPajama, outperforming the full-data pre-training within a 50B training budget, and achieving about 1.5x training efficiency and 5x data efficiency.