Distilling the Thought, Watermarking the Answer: A Principle Semantic Guided Watermark for Large Reasoning Models

2026-01-08

arXiv.org

AI Summary

This work proposes a principal semantics-guided adaptive watermarking mechanism to address the common trade-offs in existing large language model watermarking methods, which often compromise logical coherence or incur high computational overhead. The approach decouples text generation into an undisturbed reasoning phase and a watermark-embedded response phase. By extracting core semantic elements from the reasoning trajectory via salience scoring, it constructs a principal semantics vector to dynamically modulate watermark strength. This design preserves logical integrity while significantly enhancing watermark robustness and generation efficiency: it reduces text perplexity by 0.35, improves translation BLEU score by 0.164, increases mathematical accuracy by 0.67 percentage points, and boosts watermark detection AUC by 0.34%, all with negligible impact on inference latency.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Reasoning Large Language Models (RLLMs) excelling in complex tasks present unique challenges for digital watermarking, as existing methods often disrupt logical coherence or incur high computational costs. Token-based watermarking techniques can corrupt the reasoning flow by applying pseudo-random biases, while semantic-aware approaches improve quality but introduce significant latency or require auxiliary models. This paper introduces ReasonMark, a novel watermarking framework specifically designed for reasoning-intensive LLMs. Our approach decouples generation into an undisturbed Thinking Phase and a watermarked Answering Phase. We propose a Criticality Score to identify semantically pivotal tokens from the reasoning trace, which are distilled into a Principal Semantic Vector (PSV). The PSV then guides a semantically-adaptive mechanism that modulates watermark strength based on token-PSV alignment, ensuring robustness without compromising logical integrity. Extensive experiments show ReasonMark surpasses state-of-the-art methods by reducing text Perplexity by 0.35, increasing translation BLEU score by 0.164, and raising mathematical accuracy by 0.67 points. These advancements are achieved alongside a 0.34% higher watermark detection AUC and stronger robustness to attacks, all with a negligible increase in latency. This work enables the traceable and trustworthy deployment of reasoning LLMs in real-world applications.

Problem

• watermarking
• reasoning LLMs
• logical coherence
• computational cost
• semantic integrity

Innovation

• ReasonMark
• Principal Semantic Vector
• semantic-aware watermarking
• reasoning LLMs
• criticality score