As large language models(LLMs) are increasingly deployed in real-world applications, they inevitably encounter culturally sensitive scenarios that require applying cultural knowledge or values to the context in order to generate appropriate responses. However, existing research largely treats culture as static knowledge or abstract values, leaving the application of cultural norms in situated interactions underexplored. To address this gap, we first focus on the situated cultural reasoning of LLMs by proposing CuSiR, a simulation learning framework combining simulated scenario and reinforcement learning. We construct datasets covering both knowledge-based and social scenarios and conduct experiments across multiple role perspectives, instructions, settings, and models. Our results indicate that this framework effectively enhances LLMs’ ability to apply existing cultural knowledge, thereby improving their performance on tasks set in cultural scenarios. Code files are available at https://github.com/HoganZinger/Simulation_Learning

Evaluating the multilingual and multicultural capabilities of Large Language Models (LLMs) is essential for their global utility. However, current benchmarks face three critical limitations: (1) fragmented evaluation dimensions that often neglect deep cultural nuances; (2) insufficient language coverage in subjective tasks relying on low-quality machine translation; and (3) shallow analysis that lacks diagnostic depth beyond simple rankings. To address these, we introduce GaoYao, a comprehensive benchmark with 182.3k samples, 26 languages and 51 nations/areas. First, GaoYao proposes a unified framework categorizing evaluation tasks into three cultural layers (General Multilingual, Cross-cultural, Monocultural) and nine cognitive sub-layers. Second, we achieve native-quality expansion by leveraging experts to rigorously localize subjective benchmarks into 19 languages and synthesizing cross-cultural test sets for 34 cultures, surpassing prior coverage by up to 111%. Third, we conduct an in-depth diagnostic analysis on 20+ flagship and compact LLMs. Our findings reveal significant geographical performance disparities and distinct gaps between tasks, offering a reliable map for future work. We release the benchmark.

Cultural taboo safety is essential for deploying large language models (LLMs), as culturally insensitive outputs may cause offense or even social harm. However, existing cultural benchmarks primarily assess cultural knowledge or values biases, while overlooking whether LLMs can recognize and respect cultural taboos, especially when taboos are implicitly hidden in seemingly harmless questions. Besides, cultural taboos are implicit, and context-dependent, thus poss unique challenges for reliable evaluation. To address these gaps, we introduce CulShield, the first public benchmark dedicated to evaluating and improving the cultural taboo safety of LLMs. CulShield spans 77 countries and regions, and includes over 2,020 taboos. It evaluates models along both explicit knowledge and implicit behaviors.Experiments on several advanced LLMs (e.g., GPT-4o-mini, Gemini-2.5-pro) reveal a clear “knowledge-behavior gap”: models often fail to apply known taboos during interaction. We further show that variations in linguistic context can significantly affect LLMs’ cultural taboo safety. Code and data is accessible here: https://anonymous.4open.science/r/CulShield-7A0E.

With contributions from the open-source community, a vast amount of instruction tuning (IT) data has emerged. Given the significant resource allocation required by training and evaluating models, it is advantageous to have an efficient method for selecting high-quality IT data. However, existing methods for instruction data selection have limitations such as relying on fragile external APIs, being affected by biases in GPT models, or reducing the diversity of the selected instruction dataset. In this paper, we propose an industrial-friendly, expert-aligned and diversity-preserved instruction data selection method: Clustering and Ranking (CaR). CaR consists of two steps. The first step involves ranking instruction pairs using a scoring model that is well aligned with expert preferences (achieving an accuracy of 84.25%). The second step involves preserving dataset diversity through a clustering process. In our experiment, CaR selected a subset containing only 1.96% of Alpaca’s IT data, yet the underlying AlpaCaR model trained on this subset outperforms Alpaca by an average of 32.1% in GPT-4 evaluations. Furthermore, our method utilizes small models (550M parameters) and requires only 11.2% of the monetary cost compared to existing methods, making it easily deployable in industrial scenarios.