Inferring apparent personality from facial images is crucial for embodied agents in social interactions. Unlike intrinsic personality inference through conversation, this task focuses on first-impression personality perception based solely on facial appearance. Existing studies primarily target the Big Five personality model and often depend on language or multimodal inputs, leaving the effectiveness of facial cues alone in establishing meaningful personality trait associations unclear. This is particularly relevant for MBTI types, which are widely utilized and interpretable by large language models. To address this, we propose GlanceFace, an end-to-end framework for apparent personality inference that leverages vision-language models to introduce semantic priors, along with a semantic-enhanced facial representation module to capture subtle personality-related cues. Additionally, we implement an uncertainty-aware learning strategy to manage noisy and subjective annotations. Extensive experiments demonstrate strong performance on MBTI-based apparent personality benchmarks and reveal relationships between facial characteristics and perceived personality traits, highlighting its potential to support adaptive initial interaction strategies for embodied agents. The code and dataset are available at GitHub.
Blogger's Review: The GlanceFace framework proposed in this paper significantly enhances the correlation between facial features and personality inference by integrating vision and language models. It opens new avenues for personalized adaptive strategies in human-robot interactions, making it a noteworthy contribution to the field.