Personalized alignment is essential for enabling large language models (LLMs) to engage effectively in user-centric dialogue. While recent prompt-based and offline optimization methods offer preliminary solutions, they fall short in cold-start scenarios and long-term personalization due to their inherently static and shallow designs. In this work, we introduce the Reinforcement Learning for Personalized Alignment (RLPA) framework, in which an LLM interacts with a simulated user model to iteratively infer and refine user profiles through dialogue. The training process is guided by a dual-level reward structure: the Profile Reward encourages accurate construction of user representations, while the Response Reward incentivizes generation of responses consistent with the inferred profile. We instantiate RLPA by fine-tuning Qwen-2.5-3B-Instruct, resulting in Qwen-RLPA, which achieves state-of-the-art performance in personalized dialogue. Empirical evaluations demonstrate that Qwen-RLPA consistently outperforms prompting and offline fine-tuning baselines, and even surpasses advanced commercial models such as Claude-3.5 and GPT-4o. Further analysis highlights Qwen-RLPA's robustness in reconciling conflicting user preferences, sustaining long-term personalization and delivering more efficient inference compared to recent reasoning-focused LLMs. These results emphasize the potential of dynamic profile inference as a more effective paradigm for building personalized dialogue systems.

Recent advances in Large Language Models (LLMs) have spurred transformative applications in various domains, ranging from open-source to proprietary LLMs. However, jailbreak attacks, which aim to break safety alignment and user compliance by tricking the target LLMs into answering harmful and risky responses, are becoming an urgent concern. The practice of red-teaming for LLMs is to proactively explore potential risks and error-prone instances before the release of frontier AI technology. This paper proposes an agentic workflow to automate and scale the red-teaming process of LLMs through the Composition-of-Principles (CoP) framework, where human users provide a set of red-teaming principles as instructions to an AI agent to automatically orchestrate effective red-teaming strategies and generate jailbreak prompts. Distinct from existing red-teaming methods, our CoP framework provides a unified and extensible framework to encompass and orchestrate human-provided red-teaming principles to enable the automated discovery of new red-teaming strategies. When tested against leading LLMs, CoP reveals unprecedented safety risks by finding novel jailbreak prompts and improving the best-known single-turn attack success rate by up to 19.0 times.

Software languages evolve over time for various reasons, such as the addition of new features. When the language's grammar definition evolves, textual instances that originally conformed to the grammar become outdated. For DSLs in a model-driven engineering context, there exists a plethora of techniques to co-evolve models with the evolving metamodel. However, these techniques are not geared to support DSLs with a textual syntax -- applying them to textual language definitions and instances may lead to the loss of information from the original instances, such as comments and layout information, which are valuable for software comprehension and maintenance. This study explores the potential of Large Language Model (LLM)-based solutions in achieving grammar and instance co-evolution, with attention to their ability to preserve auxiliary information when directly processing textual instances. By applying two advanced language models, Claude-3.5 and GPT-4o, and conducting experiments across seven case languages, we evaluated the feasibility and limitations of this approach. Our results indicate a good ability of the considered LLMs for migrating textual instances in small-scale cases with limited instance size, which are representative of a subset of cases encountered in practice. In addition, we observe significant challenges with the scalability of LLM-based solutions to larger instances, leading to insights that are useful for informing future research.

Multimodal Large Language Models (MLLMs) have shown significant potential in surgical video understanding. With improved zero-shot performance and more effective human-machine interaction, they provide a strong foundation for advancing surgical education and assistance. However, existing research and datasets primarily focus on understanding surgical procedures and workflows, while paying limited attention to the critical role of anatomical comprehension. In clinical practice, surgeons rely heavily on precise anatomical understanding to interpret, review, and learn from surgical videos. To fill this gap, we introduce the Neurosurgical Anatomy Benchmark (NeuroABench), the first multimodal benchmark explicitly created to evaluate anatomical understanding in the neurosurgical domain. NeuroABench consists of 9 hours of annotated neurosurgical videos covering 89 distinct procedures and is developed using a novel multimodal annotation pipeline with multiple review cycles. The benchmark evaluates the identification of 68 clinical anatomical structures, providing a rigorous and standardized framework for assessing model performance. Experiments on over 10 state-of-the-art MLLMs reveal significant limitations, with the best-performing model achieving only 40.87% accuracy in anatomical identification tasks. To further evaluate the benchmark, we extract a subset of the dataset and conduct an informative test with four neurosurgical trainees. The results show that the best-performing student achieves 56% accuracy, with the lowest scores of 28% and an average score of 46.5%. While the best MLLM performs comparably to the lowest-scoring student, it still lags significantly behind the group's average performance. This comparison underscores both the progress of MLLMs in anatomical understanding and the substantial gap that remains in achieving human-level performance.

Large Language Models (LLMs) have recently emerged as powerful tools for autoformalization. Despite their impressive performance, these models can still struggle to produce grounded and verifiable formalizations. Recent work in text-to-SQL, has revealed that LLMs can be sensitive to paraphrased natural language (NL) inputs, even when high degrees of semantic fidelity are preserved (Safarzadeh, Oroo-jlooyjadid, and Roth 2025). In this paper, we investigate this claim in the autoformalization domain. Specifically, we evaluate the robustness of LLMs generating formal proofs with semantically similar paraphrased NL statements by measuring semantic and compilation validity. Using the formal benchmarks MiniF2F (Zheng, Han, and Polu 2021) and Lean 4 version of ProofNet (Xin et al. 2024), and two modern LLMs, we generate paraphrased natural language statements and cross-evaluate these statements across both models. The results of this paper reveal performance variability across paraphrased inputs, demonstrating that minor shifts in NL statements can significantly impact model outputs.

Large Language Model (LLM) agents are increasingly studied in multi-turn, multi-agent scenarios, yet most existing setups emphasize open-ended role-play rather than controlled evaluation. We introduce AsymPuzl, a minimal but expressive two-agent puzzle environment designed to isolate communication under information asymmetry. Each agent observes complementary but incomplete views of a symbolic puzzle and must exchange messages to solve it cooperatively. Using a diverse set of current-generation and open-source LLMs, we show that (i) strong models such as GPT-5 and Claude-4.0 reliably converge across puzzle sizes on the solution by sharing complete information in two turns, (ii) weaker models often ignore partner messages or over-correct their hypotheses, and (iii) feedback design is non-trivial: simple self-feedback improves success rates, while detailed joint feedback can hurt performance. These findings show that even in simple cooperative tasks, LLM communication strategies diverge and depend on the granularity of feedback signals. AsymPuzl thus provides a testbed for probing the limits of multi-turn cooperation and opens avenues for studying coordination mechanisms.

Motivated by the success of general-purpose large language models (LLMs) in software patching, recent works started to train specialized patching models. Most works trained one model to handle the end-to-end patching pipeline (including issue localization, patch generation, and patch validation). However, it is hard for a small model to handle all tasks, as different sub-tasks have different workflows and require different expertise. As such, by using a 70 billion model, SOTA methods can only reach up to 41% resolved rate on SWE-bench-Verified. Motivated by the collaborative nature, we propose Co-PatcheR, the first collaborative patching system with small and specialized reasoning models for individual components. Our key technique novelties are the specific task designs and training recipes. First, we train a model for localization and patch generation. Our localization pinpoints the suspicious lines through a two-step procedure, and our generation combines patch generation and critique. We then propose a hybrid patch validation that includes two models for crafting issue-reproducing test cases with and without assertions and judging patch correctness, followed by a majority vote-based patch selection. Through extensive evaluation, we show that Co-PatcheR achieves 46% resolved rate on SWE-bench-Verified with only 3 x 14B models. This makes Co-PatcheR the best patcher with specialized models, requiring the least training resources and the smallest models. We conduct a comprehensive ablation study to validate our recipes, as well as our choice of training data number, model size, and testing-phase scaling strategy.

Despite emerging efforts to enhance the safety of Vision-Language Models (VLMs), current approaches face two main shortcomings. 1) Existing safety-tuning datasets and benchmarks only partially consider how image-text interactions can yield harmful content, often overlooking contextually unsafe outcomes from seemingly benign pairs. This narrow coverage leaves VLMs vulnerable to jailbreak attacks in unseen configurations. 2) Prior methods rely primarily on data-centric tuning, with limited architectural innovations to intrinsically strengthen safety. We address these gaps by introducing a holistic safety dataset and benchmark, \textbf{HoliSafe}, that spans all five safe/unsafe image-text combinations, providing a more robust basis for both training and evaluation (HoliSafe-Bench). We further propose a novel modular framework for enhancing VLM safety with a visual guard module (VGM) designed to assess the harmfulness of input images for VLMs. This module endows VLMs with a dual functionality: they not only learn to generate safer responses but can also provide an interpretable harmfulness classification to justify their refusal decisions. A significant advantage of this approach is its modularity; the VGM is designed as a plug-in component, allowing for seamless integration with diverse pre-trained VLMs across various scales. Experiments show that Safe-VLM with VGM, trained on our HoliSafe, achieves state-of-the-art safety performance across multiple VLM benchmarks. Additionally, the HoliSafe-Bench itself reveals critical vulnerabilities in existing VLM models. We hope that HoliSafe and VGM will spur further research into robust and interpretable VLM safety, expanding future avenues for multimodal alignment.

Large Language Models (LLMs) demonstrate partial forecasting competence across social, political, and economic events. Y et, their predictive ability varies sharply with domain structure and prompt framing. We investigate how forecasting performance varies with different model families on real-world questions about events that happened beyond the model cutoff date. We analyze how context, question type, and external knowledge affect accuracy and calibration, and how adding factual news context modifies belief formation and failure modes. Our results show that forecasting ability is highly variable as it depends on what, and how, we ask.

As libraries explore large language models (LLMs) for use in virtual reference services, a key question arises: Can LLMs serve all users equitably, regardless of demographics or social status? While they offer great potential for scalable support, LLMs may also reproduce societal biases embedded in their training data, risking the integrity of libraries' commitment to equitable service. To address this concern, we evaluate whether LLMs differentiate responses across user identities by prompting six state-of-the-art LLMs to assist patrons differing in sex, race/ethnicity, and institutional role. We find no evidence of differentiation by race or ethnicity, and only minor evidence of stereotypical bias against women in one model. LLMs demonstrate nuanced accommodation of institutional roles through the use of linguistic choices related to formality, politeness, and domain-specific vocabularies, reflecting professional norms rather than discriminatory treatment. These findings suggest that current LLMs show a promising degree of readiness to support equitable and contextually appropriate communication in academic library reference services.