The paper introduces MIRAGE, a framework for evaluating LLMs' ability to simulate human behaviors in murder mystery games. MIRAGE uses four methods: TII, CIC, ICI and SCI to assess the LLMs' role-playing proficiency. Experiments show that even GPT-4 struggles with the complexities of the MIRAGE framework.
MBZUAI researchers introduce SocialMaze, a new benchmark for evaluating social reasoning capabilities in large language models (LLMs). SocialMaze includes six diverse tasks across social reasoning games, daily-life interactions, and digital community platforms, emphasizing deep reasoning, dynamic interaction, and information uncertainty. Experiments show that LLMs vary in handling dynamic interactions, degrade under uncertainty, but can be improved via fine-tuning on curated reasoning examples.
A new benchmark, LongShOTBench, is introduced for evaluating multimodal reasoning and tool use in long videos, featuring open-ended questions and diagnostic rubrics. The benchmark addresses the limitations of existing datasets by combining temporal length and multimodal richness, using human-validated samples. LongShOTAgent, an agentic system, is also presented for analyzing long videos, with both the benchmark and agent demonstrating the challenges faced by state-of-the-art MLLMs.
This paper introduces Arabic language integration into Vision-and-Language Navigation (VLN) in robotics, evaluating multilingual SLMs like GPT-4o mini, Llama 3 8B, Phi-3 14B, and Jais using the NavGPT framework. The study uses the R2R dataset to assess the impact of language on navigation reasoning through zero-shot sequential action prediction. Results show the framework enables high-level planning in both English and Arabic, though some models face challenges with Arabic due to reasoning limitations and parsing issues. Why it matters: This work highlights the need to improve language model planning and reasoning for effective navigation, especially to unlock the potential of Arabic-language models in real-world applications.