INT v.s. FP: A Comprehensive Study of Fine-Grained Low-bit Quantization Formats Paper • 2510.25602 • Published Oct 29 • 77
POINTS-Reader: Distillation-Free Adaptation of Vision-Language Models for Document Conversion Paper • 2509.01215 • Published Sep 1 • 50
Over-Tokenized Transformer: Vocabulary is Generally Worth Scaling Paper • 2501.16975 • Published Jan 28 • 31