The Magic Correlations: Understanding Knowledge Transfer from Pretraining to Supervised Fine-Tuning

TL;DR

This paper investigates how language model capabilities and confidence metrics transfer from pretraining to supervised fine-tuning, revealing complex dynamics that inform better benchmark selection and data curation strategies.

Contribution

It provides a comprehensive analysis of transfer reliability across different capabilities, benchmarks, and scales, highlighting the nuanced behavior of accuracy and confidence during fine-tuning.

Findings

Transfer reliability varies across capabilities and benchmarks.

Accuracy and confidence exhibit distinct scaling behaviors.

Benchmark selection impacts transfer effectiveness.

Abstract

Understanding how language model capabilities transfer from pretraining to supervised fine-tuning (SFT) is fundamental to efficient model development and data curation. In this work, we investigate four core questions: RQ1. To what extent do accuracy and confidence rankings established during pretraining persist after SFT? RQ2. Which benchmarks serve as robust cross-stage predictors and which are unreliable? RQ3. How do transfer dynamics shift with model scale? RQ4. How well does model confidence align with accuracy, as a measure of calibration quality? Does this alignment pattern transfer across training stages? We address these questions through a suite of correlation protocols applied to accuracy and confidence metrics across diverse data mixtures and model scales. Our experiments reveal that transfer reliability varies dramatically across capability categories, benchmarks, and scales -- with accuracy and confidence exhibiting distinct, sometimes opposing, scaling dynamics. These findings shed light on the complex interplay between pretraining decisions and downstream outcomes, providing actionable guidance for benchmark selection, data curation, and efficient model development.