TL;DR
A hybrid approach that combines contextual embeddings from RoBERTa with handcrafted stylometric features to capture both high-level semantic patterns and low-level stylistic cues enables robust generalization across heterogeneous text types is adopted.
A hybrid detection framework which combines semantically deep embeddings from the RoBERTa transformer with a set of carefully designed language statistics and linguistic statistics and shows excellent resistance to the surface-level adversarial paraphrasing strategy.
Authorship verification (AV) assumes that an author's writing style remains sufficiently stable to distinguish it from that of other writers. In practice, however, this assumption is challenged by distribution shifts caused by changes in genre, time, and AI-assisted writing. Existing AV benchmarks typically study these factors in isolation and focus predominantly on English, limiting our understanding of model robustness under realistic conditions. We introduce AVShift, the first German benchmark for systematically evaluating AV under multiple distribution shifts. AVShift comprises over 150,000 text pairs spanning three genres and 21 years, enabling controlled evaluation of cross-genre, temporal, and AI-era shifts within a unified framework. We benchmark representative feature-based, embedding-based, and LLM-based approaches. Our experiments show that fine-tuned LLMs generalize best across genres and benefit substantially from stylistically diverse training data. We further demonstrate that temporal drift is one of the strongest factors affecting AV, with performance degrading significantly as the time gap between documents increases. In contrast, we find no evidence of a measurable AI-era distribution shift within AVShift. Finally, our feature analysis reveals stylistic features that remain stable across genres, while their relative importance varies depending on the specific genre transition. We release AVShift and our code for future research.
This work investigates whether large language models (LLMs) can automatically adapt template-based bias datasets to specific domains using zero-shot prompting and shows that domain-adapted templates capture real-world bias patterns more faithfully than standard templates.
The XAI analysis reveals that interpreting the VAE-based model’s decision uncovers stylistic differences between the two text types beyond superficial lexical and syntactic features, demonstrating the potential of the XAI DL framework in other NLP studies that aim to analyze style.
This work compares data-driven knowledge distillation (DDKD) against zero-shot inference and fine-tuning on out-of-domain D2T data, and introduces structure-preserving augmentation via structural subsampling and perturbation in cross-domain D2T generation.
Author verification is a fundamental task in natural language processing with potential applications in digital forensics, copyright dispute resolution, and social media account linkage. While existing deep pre-trained models achieve high accuracy, they incur substantial computational costs, high inference latency, and poor interpretability, limiting their deployment in resource-constrained environments. To address this gap, a lightweight and interpretable author verification method is proposed that relies solely on hand-crafted stylistic features and classical machine learning classifiers, requiring no GPU acceleration. Specifically, three complementary feature sets are extracted per text: 37 punctuation features, 22 text-style features, and an 800-dimensional character-level TF-IDF vector. For each text pair, we compute the absolute difference and element-wise product for each feature group separately, and we also concatenate all resulting vectors to form the final classification representation. On a public review dataset under a strict author-level split, the proposed method achieves competitive verification performance with significantly improved training and inference efficiency. Notably, while the lightweight neural architectures evaluated in this study yield marginal accuracy gains, they incur substantially longer inference and training times, validating the efficiency-performance trade-off of our approach. Ablation studies confirm the contribution of each feature group, with text statistics being the most influential. The proposed method offers an efficient, transparent, and easily deployable solution for author verification in resource-limited settings and provides empirical evidence for integrating traditional stylistic features with lightweight models.