Animal assisted intervention has been one of the effective natural therapeutic approaches, especially for individuals with autism. To increase the accessibility and reduce extra burden of care for the animals, Robotic-Animal Assisted Interventions (RAAI) has been proposed. However, the lack of natural behaviors is one of the key factors in limiting the feasibility with the current technology. This late-breaking report aims to build natural behavior models with data-driven approach, utilizing latest development of large language models (LLMs) to effectively analyze the data and build natural language-based models. Due to the reliance on LLM in this study, a key limitation is the lack of continuity in understanding. Frame images, as static representations, may not fully capture temporal dynamics. Future studies could address this limitation by integrating 3D-pose analysis, which would improve both continuity and contextual understanding.

Recent advances in enhancing the reasoning ability of Large Language Models (LLMs) have been remarkably successful. LLMs trained with Reinforcement Learning (RL) for reasoning demonstrate strong performance in challenging tasks such as mathematics and coding, even with relatively small model sizes. However, despite these impressive improvements in task accuracy, the assessment of creativity in LLM generations has been largely overlooked in reasoning tasks, in contrast to writing tasks. The lack of research on creativity assessment in reasoning primarily stems from two challenges: (1) the difficulty of defining the range of creativity, and (2) the necessity of human evaluation in the assessment process. To address these challenges, we propose CLAWS, a novel method that defines and classifies mathematical solutions into Typical, Creative, and Hallucinated categories without human evaluation, by leveraging attention weights across prompt sections and output. CLAWS outperforms five existing white-box detection methods—Perplexity, Logit Entropy, Window Entropy, Hidden Score, and Attention Score—on five 7–8B math RL models (DeepSeek, Qwen, Mathstral, OpenMath2, and Oreal). We validate CLAWS on 4,545 math problems collected from 181 math contests (A(J)HSME, AMC, AIME). Our code is available at https://github.com/kkt94/CLAWS.