Predictable Scaling (PS) is a framework that uses small‑scale training proxies to forecast full‑scale model performance, reducing the need for expensive GPU‑hourly runs. By applying composable scaling functions and curve fitting with uncertainty estimates, PS can rank candidate changes accurately across different hypothesis types. Experiments on two ranking models show PS achieves 79–88% compute savings while maintaining correct ranking of actionable hypotheses.
arXiv:2610.27789v1 Announce Type: new Abstract: Empirical scaling laws relate validation loss to model size, training exposure and compute, providing a quantitative basis for allocating resources. Th...
arXiv:2604. 22753v2 Announce Type: replace Abstract: Scaling laws are used to plan multi-million-dollar training runs, but fitting those laws can itself cost millions.
The paper introduces Power‑Law Entropy Search (PLES), a computational‑cost‑aware acquisition function that uses multi‑fidelity Bayesian optimization to efficiently estimate optimal hyperparameter scaling laws for large language model training. PLES focuses on reducing the overall uncertainty of scaling law estimates rather than optimizing a single objective, selecting configurations that maximize uncertainty reduction per unit computational cost. Experiments on synthetic benchmarks, surrogate models, and real LLM pre‑training runs show that PLES converges to accurate scaling laws using less than one‑tenth of the computational budget required by conventional grid search and other baselines.
The paper proposes a framework to reduce the cost of constructing scaling laws for large foundation models by treating data collection as a Bayesian optimization problem. It shows that expanding the compute budget progressively and augmenting observed configurations with surrogate-fantasized evaluations can recover a broad experimental grid, enabling accurate scaling law fitting without training every configuration. This approach can achieve computational savings of up to 10–100× compared to a full dense grid.
arXiv:2605. 29283v2 Announce Type: replace-cross Abstract: Recent physics foundation models claim general spatiotemporal forecasting ability, yet their evaluations often collapse performance into a single average score under a fixed training distribution.
arXiv:2602. 15327v2 Announce Type: replace-cross Abstract: Machine learning model performance improvements tend to arise from competition and application.
arXiv:2605.30018v3 Announce Type: replace-cross Abstract: Large language models (LLMs) frequently achieve impressive scores on standardized benchmarks, yet accuracy alone offers a limited view of the...
Scaling laws promised cost-effective experiments; six years later, they have yet to fully deliver. Instead, researchers have found them unreliable at small scales (starting at 4M parameters) and concluded that sizable models cannot be avoided.
arXiv:2607. 23377v1 Announce Type: cross Abstract: The largest machine learning models in particle physics are also the most expensive to train, yet the return on scaling a given architecture cannot be estimated before that compute is spent.
arXiv:2603. 24226v4 Announce Type: replace-cross Abstract: Recent advances in Large Language Models (LLMs) have inspired a surge of scaling research in industrial search, advertising, and recommendation systems.
The largest machine learning models in particle physics are also the most expensive to train, yet the return on scaling a given architecture cannot be estimated before that compute is spent. Scaling laws have been fit for jets, but none has yet been shown to predict the performance of models it was not fit on.
PhysicsBench is a unified benchmark and leaderboard that evaluates both generative and predictive AI models for engineering design and simulation under a single standardized procedure. It covers seven generation and prediction tasks across 1D, 2D, and 3D domains, ranking 66 models on nine industrial‑scale CAD/CFD/FEA datasets and public references, expanded into 28 configurations. The evaluation uses realistic, limited data scales and a common metric suite that captures geometric fidelity, physical‑field accuracy, and engineering‑specific validity, with rankings derived via a PageRank‑based dominance graph and a separate efficiency view.