Large language models demonstrate exceptional capabilities but raise critical safety concerns due to toxic outputs causing potential harm to users. Prompt optimization has emerged as a practical detoxification solution because it operates at the input level and can be applied to black-box or API-based models without parameter access. However, existing methods treat each task in a stateless manner without leveraging prior optimization experiences, which results in either ineffective detoxification or over-sanitized prompts that lose original intent in the vast search space. To address this, we propose the Memory-INformed Detoxification (MIND) framework that harnesses historical optimization experiences to enhance detoxification effectiveness through Memory Retrieval (MR) for retrieving relevant past cases and Memory Maintenance (MM) for maintaining a representative memory pool. We formulate MR and MM as optimization problems, prove their NP-hardness, and develop two approximation algorithms with constant approximation ratios for these problems. Experiments across multiple datasets and model configurations show that MIND outperforms existing baselines in both toxicity reduction and response utility.

Personalizing large language models requires adapting model behavior to individual users while preserving robustness and deployment-scale efficiency. Existing approaches typically personalize LLMs either at the input level, by retrieving user histories or constructing profile prompts, or at the parameter level, by maintaining user-specific parameter-efficient modules. The former makes personalization sensitive to retrieval quality and prompt design, whereas the latter incurs storage and maintenance costs that grow with the user population. To address these limitations, we propose TAP-PER (Temporal Attentive Prefix for PERsonalization), a prefix-based framework that encodes user preferences as learnable representations, avoiding the serialization of user histories into prompts and replacing heavy per-user adapters with lightweight user-state prefix embeddings. Inspired by personalized recommendation systems, TAP-PER decomposes user modeling into user-state and query-conditioned components, and incorporates temporal signals to capture the evolving nature of user interests. Experiments on six LaMP tasks show that TAP-PER consistently outperforms prompt-based and model-based baselines across classification, rating, and generation settings. Moreover, TAP-PER uses 130× fewer per-user parameters than OPPU and roughly half the total parameter footprint of PER-PCS at the 1,000-user scale, demonstrating scalable personalization without prompt-serialized histories or heavy per-user adapters.

Automatic harmfulness metrics and LLM-based judges are central to LLM safety assessment, yet differences in evaluation criteria, methodologies, and output formats may yield inconsistent judgments for the same model response. The lack of systematic research on these evaluators undermines the credibility and consistency of LLM safety assessments. To bridge this gap, we introduce HarmMetric Eval, a benchmark for both overall and fine-grained assessment of harmfulness evaluators. We construct a dataset of representative harmful prompts paired with diverse harmful responses and non-harmful counterparts across multiple categories. We propose a flexible scoring mechanism that rewards metrics for correctly ranking harmful responses above non-harmful ones, making it applicable to nearly all metrics and judges with varying output formats and scoring scales. Using HarmMetric Eval, we find that conventional reference-based metrics such as ROUGE and METEOR can outperform existing LLM-based judges in fine-grained harmfulness evaluation, challenging prevailing assumptions about LLM superiority. Fine-grained analysis reveals that LLM judges tend to overestimate irrelevant and useless responses. Furthermore, we build a new harmfulness judge by incorporating fine-grained criteria into its prompt template and leveraging reference-based metrics to fine-tune its base LLM. The resulting judge demonstrates leading performance on HarmMetric Eval and two external human-labeled datasets.

Cultural reasoning lies at the core of many emerging AI applications, yet existing benchmarks provide limited insight into how well models understand and contextualize cultural artifacts. We introduce ExhibitionBench, the first benchmark that frames exhibition curation as a high-level evaluation testbed, comprising 300 institution-derived exhibition and thematic collection records, 23,658 artifacts, and three interconnected tasks that evaluate complementary aspects of cultural understanding and reasoning, translating them into user-perceivable utilities. Extensive experiments across 29 systems reveal a significant gap between frontier LLMs and expert-level competence, particularly on tasks that require collection-level coherence and thematic understanding. Performance gains from scaling, prompting, and multimodal inputs are limited and highly model-dependent, whereas tailored model optimizations can equip open-weight models with better capabilities that support on-premise deployment. These findings highlight the difficulty of reasoning over cultural collections and provide a practical benchmark for developing more culturally grounded and reliable AI systems. The full project is released at https://github.com/mengze-hong/ExhibitionBench.