Social-deduction games such as Werewolf are increasingly used to evaluate LLM agents, but existing evaluations often rely on final game outcomes. We propose a belief-shift evaluation benchmark in Werewolf for analyzing communication skills through belief updating. Using LLM-played games, we annotate suspicion and accusation messages and measure how an observing village-side model’s beliefs change after each message. We evaluate 40 open-weight LLM configurations on 1,224 annotated messages. Our results show that larger models better distinguish true wolves from villagers based on game history, but accusations still strongly influence their beliefs. Models become more suspicious of the accused target and less suspicious of the accuser, especially when the accuser is trusted, even if the accuser is wolf-aligned. Larger models better resist accusations from accusers they already distrust. Overall, our findings suggest that current open-weight LLMs up to 120B parameters still struggle to integrate accusation content with source trust in strategic communication. Our benchmark and code are available at https://rlg.iis.sinica.edu.tw/papers/werewolf-accusation-benchmark.

As Large Language Models (LLMs) increasingly encourage users to disclose personal profiles for tailored assistance, measuring their political alignment becomes increasingly important. However, many existing benchmarks for assessing political behavior rely on closed-ended questions and do not fully capture how a model’s stance may adapt to user-provided context during interaction. We introduce a framework that disentangles two distinct triggers of political sycophancy: opinion (aligning with explicit narratives) and identity (stereotyping based on demographic labels). Using 450 manually-checked political dilemmas as controlled probes, we evaluate 13 instruction-tuned LLMs. We uncover a dissociation: a model’s susceptibility to explicit opinions does not necessarily predict its susceptibility to identity cues, and vice versa. When both signals are present, their effects are generally sub-additive rather than simply additive. Additionally, system-level personas primarily shift a model’s baseline stance while having limited effect on the stance shift caused by user opinion or identity. Ultimately, our results suggest that LLM political stance is interactively and steerably vulnerable rather than being a fixed trait, highlighting how personalization may amplify identity- or opinion-conditioned shifts in the model’s behaviors.

We introduce TaiwanVQA, a novel visual question answering benchmark designed to evaluate vision language models’ (VLMs) ability to recognize and reason about Taiwan-specific multimodal content.TaiwanVQA comprises 2,000 image-question pairs covering diverse topics relevant to Taiwanese culture and daily life. We categorize the questions into recognition and reasoning tasks, further sub-classifying reasoning questions based on the level of external knowledge required. We conduct extensive experiments on state-of-the-art VLMs, including GPT-4o, Llama-3.2, LLaVA, Qwen2-VL, and InternVL2 models. Our findings reveal significant limitations in current VLMs when handling culturally specific content. The performance gap widens between recognition tasks (top score 73.60%) and reasoning tasks (top score 49.80%), indicating challenges in cultural inference and contextual understanding.These results highlight the need for more culturally diverse training data and improved model architectures that can better integrate visual and textual information within specific cultural contexts. By providing TaiwanVQA, we aim to contribute to the development of more inclusive and culturally aware AI models, facilitating their deployment in diverse real-world settings. TaiwanVQA can be accessed on our GitHub page.