国内数字创意领军者，AI赋能未来可期

Mathematical reasoning in visual contexts Natural image understanding 77.8% VQAv2 0-ahot Gemini Ultra (pixel only") O-shot GPT-4V 82.3% 78.0% TextVQA OCR on natural images 0-shot Gemini Ultra (pixel only") O-shot GPT-4V Document understanding 90.9% 88.4% DocVQA 0-shot O-shot Gemini Ultra (pixel only*) Infographic understanding 80.3% 75.1% Infographic VQA 4-shot DeepMind Flamingo Video question answering 54.7% 46.3% Perception Test 0-shot Gemini Ultra O-shot SeViLA 图表 21 Gemini 根据指示教工作人员"鸭子"的普通话发音并解释汉语声调 资料来 ...