AGENTIC-SYSTEMSJuly 24, 20261 min read

Marking the Wrong Symptoms: Evaluating LLM Watermarks in Medical Texts

A new comprehensive study reveals that applying LLM watermarking schemes in the medical domain leads to severe performance degradation, inducing lexical corruption, hallucinations, and misinterpretations in clinical reasoning.

Computer Science > Artificial Intelligence

Title:Marking the Wrong Symptoms: Evaluating LLM Watermarks in Medical Texts

View PDF HTML (experimental)Abstract:Large language models (LLMs) are increasingly integrated into clinical workflows, stressing the need for reliable traceability of model-generated output with watermarking. Yet, most watermarks are evaluated on general-purpose benchmarks, leaving domains like medicine, where small token-level perturbations can result in significant semantic changes, underexplored. In this work, we present the first rigorous study of how LLM watermarks affect medical performance, benchmarking 5 watermarking schemes across 11 LLMs and 7 VLMs on various tasks spanning unimodal and multimodal clinical reasoning. Importantly, we complement existing evaluations by introducing a human-expert-validated pipeline for systematically auditing medical reasoning quality, terminological precision, and induced hallucinations. Our results reveal that watermarking can induce substantial degradation across multiple failure modes, including lexical corruption, hallucinated terminology, and amplified misattribution or omission of image findings. Notably, we find that the absence of domain-specific analyses, combined with aggregate metrics that miss failures inherent to clinical text, can systematically obscure practical watermark-induced degradations. Our findings establish domain-specific evaluation as a prerequisite for the safe deployment of watermarked models in medicine, where current benchmarks can otherwise mask clinically consequential failures.

Source: arXiv cs.AI Recent

More in this category

agentic-systems

InferenceBench: A Benchmark for Open-Ended LLM Inference Optimization by AI Agents

InferenceBench is a new benchmark evaluating AI agents on open-ended LLM inference optimization tasks under realistic constraints. While agents outperform default setups, qualitative analysis reveals key limitations in their strategic exploration capabilities.

agentic-systems

PersonaTrail: Benchmarking Personalized Web Agents through Browsing Trails

Researchers have introduced PersonaTrail, a novel benchmark evaluating personalized web agents through realistic browsing histories. Alongside it, the PACMem memory framework structures raw browsing data to significantly boost agent performance in inferring user preferences.

agentic-systems

Beyond Liars' Bench: The Impact of Lie Typology, Depth, and Sparsity on Deception Detection in LLMs

A new study reveals that deception detection in LLMs heavily depends on lie typology and training data structure, highlighting why current detection probes struggle in out-of-domain scenarios.

agentic-systems

Benchmarking Large Language Models on Multi-Sensor Physical Hazard Assessment

A new benchmark evaluates five leading large language models on multi-sensor physical hazard assessments, revealing a critical vulnerability when evaluating combined sensor risks below individual thresholds. Despite near-perfect accuracy on single-sensor breaches, all models failed to issue precautionary signals for multi-sensor hazards.

agentic-systems

DC-Leap: Training-Free Acceleration of dLLMs via Draft-Guided Contiguous Leaping Decoding

Researchers have introduced DC-Leap, a training-free framework that accelerates inference in Diffusion Large Language Models (dLLMs) by up to 105x when combined with KV-Cache while maintaining generation quality.

agentic-systems

PlanE: Meta Planning of Data, Tuning, and Inference for Extractive-based LLMs

Researchers introduce PlanE, a novel framework featuring a Data-Tuning-Inference (DTI) planner designed to optimize data, tuning, and inference processes for extractive-based LLMs while reducing annotation costs.