FLOP-Efficient Training: Early Stopping Based on Test-Time Compute Awareness

2026-01-04

arXiv.org

AI Summary

This work addresses the high computational cost of large language model training by proposing a test-time computation (TTC)-aware training strategy that explicitly incorporates TTC into early stopping decisions. For the first time, the approach jointly optimizes intermediate checkpoints and TTC configurations to achieve a synergistic trade-off between training and inference compute, avoiding exhaustive search. An efficient method for evaluating TTC configurations is introduced, along with a breakeven boundary that characterizes the balance between training and inference FLOPs to guide resource allocation. Experiments demonstrate that the proposed method reduces training FLOPs by up to 92% across multiple tasks while maintaining or even improving model accuracy, substantially accelerating model deployment and iteration cycles.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search
• Computer Vision: Large Vision Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Scaling training compute, measured in FLOPs, has long been shown to improve the accuracy of large language models, yet training remains resource-intensive. Prior work shows that increasing test-time compute (TTC)-for example through iterative sampling-can allow smaller models to rival or surpass much larger ones at lower overall cost. We introduce TTC-aware training, where an intermediate checkpoint and a corresponding TTC configuration can together match or exceed the accuracy of a fully trained model while requiring substantially fewer training FLOPs. Building on this insight, we propose an early stopping algorithm that jointly selects a checkpoint and TTC configuration to minimize training compute without sacrificing accuracy. To make this practical, we develop an efficient TTC evaluation method that avoids exhaustive search, and we formalize a break-even bound that identifies when increased inference compute compensates for reduced training compute. Experiments demonstrate up to 92\% reductions in training FLOPs while maintaining and sometimes remarkably improving accuracy. These results highlight a new perspective for balancing training and inference compute in model development, enabling faster deployment cycles and more frequent model refreshes. Codes will be publicly released.

Problem

• FLOP-Efficient Training
• Early Stopping
• Test-Time Compute
• Training-Compute Reduction
• Large Language Models

Innovation

• TTC-aware training
• early stopping
• training FLOPs reduction
• test-time compute
• compute trade-off