Jul 2026
FAS-R1: A Unified Multi-Task MLLM for Reasoning Face Anti-Spoofing
FAS-R1 is proposed, a two-stage reasoning-oriented MLLM framework for unified FAS prediction, covering authenticity classification, attack-type recognition and spoof-region localization, and outperforms the compared systems in cross-domain authenticity generalization and answer-and-rationale quality.
Hongyang Wang, Yi-Chen Shi, Hongrui Li et al.
· arXiv.org · 0 citations