Spurious Correlation Learning in Preference Optimization: Mechanisms, Consequences, and Mitigation via Tie Training

TL;DR

This paper analyzes how preference learning methods induce reliance on spurious correlations, leading to issues like sycophancy, and proposes tie training as a mitigation strategy validated on models including large language models.

Contribution

It provides a unified theoretical framework for understanding spurious correlation learning in preference optimization and introduces tie training to mitigate this problem.

Findings

Preference learning induces reliance on spurious features via mean bias and correlation leakage.

More data from the same distribution does not reduce spurious dependence.

Tie training reduces spurious learning without harming causal feature learning.

Abstract

Preference learning methods such as Direct Preference Optimization (DPO) are known to induce reliance on spurious correlations, leading to sycophancy and length bias in today's language models and potentially severe goal misgeneralization in future systems. In this work, we provide a unified theoretical analysis of this phenomenon, characterizing the mechanisms of spurious learning, its consequences on deployment, and a provable mitigation strategy. Focusing on log-linear policies, we show that standard preference-learning objectives induce reliance on spurious features at the population level through two channels: mean spurious bias and causal--spurious correlation leakage. We then show that this reliance creates an irreducible vulnerability to distribution shift: more data from the same training distribution fails to reduce the model's dependence on spurious features. To address this, we propose tie training, a data augmentation strategy using ties (equal-utility preference pairs) to introduce data-driven regularization. We demonstrate that this approach selectively reduces spurious learning without degrading causal learning. Finally, we validate our theory on log-linear models and provide empirical evidence that both the spurious learning mechanisms and the benefits of tie training persist for neural networks and large language models.