Automatic Configuration of LLM Post-Training Pipelines

TL;DR

AutoPipe is a budget-aware framework that efficiently configures LLM post-training pipelines by learning from historical data and guiding Bayesian optimization, significantly reducing computational costs while maintaining high performance.

Contribution

AutoPipe introduces a two-stage, dataset-conditioned approach combining offline learning and online optimization for cost-effective LLM post-training configuration.

Findings

AutoPipe outperforms offline baselines in biomedical reasoning tasks.

AutoPipe achieves comparable performance to online HPO methods with less than 10% of the computational cost.

The framework effectively reduces expensive evaluations through early-stopping and predictive scoring.

Abstract

LLM post-training pipelines that combine supervised fine-tuning and reinforcement learning are difficult to configure under realistic compute budgets: the configuration space is high-dimensional and heterogeneous, stages are strongly coupled, and each end-to-end evaluation is expensive. We propose AutoPipe, a budget-aware two-stage framework for configuration selection in LLM post-training. Offline, AutoPipe learns a dataset-conditioned learning-to-rank surrogate from historical runs, capturing within-dataset preferences and providing transferable guidance toward promising regions of the configuration space. Online, for a new dataset, AutoPipe uses the offline guidance to steer Bayesian optimization and models dataset-specific deviations with a Gaussian-process residual surrogate. To reduce evaluation cost, each trial is early-stopped and scored by a learned predictor that maps early training signals to a low-cost proxy for final post-training performance. Experiments on biomedical reasoning tasks show that AutoPipe consistently outperforms offline-only baselines and achieves comparable performance with the strongest online HPO baselines while using less than 10\% of their computational cost.