Vision Braille opens 459,512 translation pairs

Vision Braille released 459,512 Chinese-to-Braille and Braille-to-Chinese examples, combining 379,152 audited rows with 80,360 newly built passage-level rows, according to its Hugging Face dataset card. The corpus spans general, mathematical, medical, traditional-medicine, and pathology text, but its license is marked “other.” For teams watching Chinese AI compete on more than model scale, the practical opportunity is a rare accessibility corpus ready for fine-tuning; the practical constraint is to clear provenance and commercial-use rights before shipping it.