[2605.11134] Spurious Correlation Learning in Preference Optimization: Mechanisms, Consequences, and Mitigation via Tie Training

Abstract:Preference learning methods like 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.

# link_14xcovv7y9d.pdf ## Metadata - PDFFormatVersion=1.7 - IsLinearized=false - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - Author=Christian Moya; Alex Semendinger; Guang Lin; Elliott Thornley - Creator=arXiv GenPDF (tex2pdf:a6404ea) - Custom.DOI=https://doi.org/10.48550/arXiv.2605.11134 - Custom.License=http://creativecommons.org/licenses/by-nc-nd/4.0/ - Custom.PTEX.Fullbanner=This is pdfTeX, Version 3.141592653-2.6-1.40.28 (TeX Live 2025) kpathsea version 6.4.1 - Custom.arXivID=https://arxiv.org/abs/2605.11134v2 - Producer=pikepdf 8

saved by

• Elliott Thornley

related reading

• [2305.18290] Direct Preference Optimization: Your Language Model is Secretly a Reward Modelarxiv.org
• [2605.01172] A Theory of Generalization in Deep Learningarxiv.org
• [2602.05910] Chunky Post-Training: Data Driven Failures of Generalizationarxiv.org
• [2604.00626] A Survey of On-Policy Distillation for Large Language Modelsarxiv.org
• [2507.06187] The Delta Learning Hypothesis: Preference Tuning on Weak Data can Yield Strong Gainsarxiv.org
• [2503.10990] Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibriumarxiv.org
• [2606.04071] Covert Influence Between Language Modelsarxiv.org
• [2605.12715] Scaling Laws for Mixture Pretraining Under Data Constraintsarxiv.org
• [2602.19362] LLMs Can Learn to Reason Via Off-Policy RLarxiv.org
• 2308.03958arxiv.org
• Distortion of AI Alignment: Does Preference Optimization Optimize for Preferences?arxiv.org
• [2605.15220] Always Learning, Always Mixing: Efficient and Simple Data Mixing All The Timearxiv.org