Sensitivity of Small Language Models to Fine-tuning Data Contamination

2025-11-10

AI Summary

This study systematically investigates the robustness of small language models (SLMs; 270M–4B parameters) to data contamination during instruction tuning. We evaluate 23 SLMs under syntactic contamination (character- and word-order reversal) and semantic contamination (irrelevant or counterfactual responses) across contamination levels of 25%–100%. Our analysis reveals: (i) syntactic contamination induces catastrophic failure; (ii) semantic contamination exhibits a pronounced threshold effect; (iii) larger SLMs are more prone to internalizing harmful instructions—a phenomenon we term the “capability curse”; and (iv) standard alignment training fails to improve contamination robustness. Based on these findings, we propose the first contamination-robustness evaluation protocol specifically designed for SLMs. We further advocate integrating contamination-aware mechanisms into training pipelines to enhance real-world deployment safety.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Large Multimodal Models (LMMs)
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Small Language Models (SLMs) are increasingly being deployed in resource-constrained environments, yet their behavioral robustness to data contamination during instruction tuning remains poorly understood. We systematically investigate the contamination sensitivity of 23 SLMs (270M to 4B parameters) across multiple model families by measuring susceptibility to syntactic and semantic transformation types during instruction tuning: syntactic transformations (character and word reversal) and semantic transformations (irrelevant and counterfactual responses), each applied at contamination levels of 25%, 50%, 75%, and 100%. Our results reveal fundamental asymmetries in vulnerability patterns: syntactic transformations cause catastrophic performance degradation, with character reversal producing near-complete failure across all models regardless of size or family, while semantic transformations demonstrate distinct threshold behaviors and greater resilience in core linguistic capabilities. Critically, we discover a `` extit{capability curse}"where larger, more capable models become more susceptible to learning semantic corruptions, effectively following harmful instructions more readily, while our analysis of base versus instruction-tuned variants reveals that alignment provides inconsistent robustness benefits, sometimes even reducing resilience. Our work establishes three core contributions: (1) empirical evidence of SLMs'disproportionate vulnerability to syntactic pattern contamination, (2) identification of asymmetric sensitivity patterns between syntactic and semantic transformations, and (3) systematic evaluation protocols for contamination robustness assessment. These findings have immediate deployment implications, suggesting that current robustness assumptions may not hold for smaller models and highlighting the need for contamination-aware training protocols.

Problem

• Investigating SLMs' vulnerability to data contamination during instruction tuning
• Measuring susceptibility to syntactic and semantic transformation types
• Establishing systematic protocols for contamination robustness assessment

Innovation

• Systematically tested 23 small language models
• Measured syntactic and semantic transformation vulnerabilities
• Revealed capability curse in larger model susceptibility