As large language models (LLMs) increasingly rely on web-scale corpora, membership inference attacks (MIAs) have become essential tools for auditing unauthorized data usage. However, traditional token-level MIAs fail to generalize to large reasoning models (LRMs) due to data distribution shifts introduced during fine-tuning. While recent studies address this by encoding reasoning traces into latent semantic spaces, they sacrifice the interpretability strictly required for real-world copyright verification. In this paper, we demonstrate that an LRM’s internal memory states are explicitly leaked through its stylistic word choices. Our analysis reveals a distinct lexical dichotomy: non-member traces tend toward a "guessing-like" tone, whereas member traces exhibit a "remember-like" style. Motivated by this interesting phenomenon, we propose TraceMIA, an interpretable MIA framework that audits training corpora by computing membership scores based on the relative frequencies of discriminative cue words. Extensive empirical evaluations across representative LRMs and diverse datasets demonstrate that TraceMIA achieves robust detection efficacy while providing the transparent, human-readable evidence critical for practical auditing scenarios.

Although large reasoning models (LRMs) exhibit exceptional mathematical reasoning capabilities on clean inputs, their reasoning accuracy drops substantially in the presence of character-level noise such as typographical errors. Critically, their confidence estimates fail to reflect the corresponding decline in reasoning accuracy. While confidence calibration offers a principled solution, existing methods predominantly target clean inputs, leaving noisy scenarios largely unexplored. To address this gap, we propose DisCal (Distribution-aware Calibration), a confidence calibration framework for character-level noisy inputs. DisCal extracts uncertainty signals from both the empirical answer distribution and the model’s predictive distribution, and integrates them via a learned calibrator to produce well-calibrated confidence. Experiments across multiple mathematical reasoning benchmarks demonstrate that DisCal consistently outperforms existing calibration methods under noisy inputs, reducing Expected Calibration Error (ECE) by up to 39.21% and improving Area Under the Receiver Operating Characteristic Curve (AUROC) by up to 31.44%.

Large language models (LLMs) have achieved groundbreaking progress in Natural Language Processing (NLP). Despite the numerous advantages of LLMs, they also pose significant safety risks. Self-evaluation mechanisms have gained increasing attention as a key safeguard to ensure safe and controllable content generation. However, LLMs often exhibit overconfidence, which seriously compromises the accuracy of safety self-evaluation. To address this challenge, we propose SafeConf, a method to enhance the safety self-evaluation capability of LLMs through confidence calibration. The method performs semantic mutations on the original safety evaluation questions and adopts a self-consistency strategy to quantify confidence based on answer accuracy on the mutated questions. Finally, these confidence scores are used to construct a dataset for fine-tuning. We conducte experiments on both Chinese and English datasets. The results show that SafeConf improves self-evaluation accuracy by an average of 5.86% and 7.79% over the state-of-the-art baseline methods on Qwen2.5-7B-Instruct and Llama3-8B-Instruct models, respectively, without affecting the general capabilities of the models.

Extractive summarization for legal documents aims to automatically extract key sentences from legal texts to form concise summaries. Recent studies have explored diffusion models for extractive summarization task, showcasing their remarkable capabilities. Despite these advancements, these models often fall short in effectively capturing and leveraging the specialized legal terminology crucial for accurate legal summarization. To address the limitation, this paper presents a novel term-guided diffusion model for extractive summarization of legal documents, named TermDiffuSum. It incorporates legal terminology into the diffusion model via a well-designed multifactor fusion noise weighting schedule, which allocates higher attention weight to sentences containing a higher concentration of legal terms during the diffusion process. Additionally, TermDiffuSum utilizes a re-ranking loss function to refine the model’s selection of more relevant summaries by leveraging the relationship between the candidate summaries generated by the diffusion process and the reference summaries. Experimental results on a self-constructed legal summarization dataset reveal that TermDiffuSum outperforms existing diffusion-based summarization models, achieving improvements of 3.10 in ROUGE-1, 2.84 in ROUGE-2, and 2.89 in ROUGE-L. To further validate the generalizability of TermDiffuSum, we conduct experiments on three public datasets from news and social media domains, with results affirming the scalability of our approach.