刘 佳祥Liu Jiaxiang
刘佳祥，机器学习研究员，毕业后加入百度，从事自然语言处理与大模型研发至今。
他主导了第一批大语言模型上线百度搜索的检索与排序系统，是该方向的早期推动者之一。凭借在自然语言理解领域的系统性工作，他的团队在 SuperGLUE Benchmark 评测中登顶世界排名第一，并因此获得百度最高奖。
此后，他深度参与 ERNIE 系列大模型的研发，工作横跨文本理解、多模态理解与多模态生成三个方向。相关成果发表于 AAAI、CVPR、KDD 等顶级学术会议，论文引用逾千次。
从 CNN、LSTM 的时代，到 BERT 的兴起，再到以 ChatGPT 为标志的生成式 AI 浪潮，刘佳祥是这一轮技术跃迁的全程参与者。当下，他正全力投入生成式 AI 的前沿探索, 开发了消费级显卡上可以运行的文生图模型 ERNIE-Image。
工作之外，他热衷摄影，偏爱克制、安静的画面语言。亦是广泛的听乐者，涉猎多元风格。
Liu Jiaxiang is a machine learning researcher who joined Baidu after graduation, where he has been working on natural language processing and large model R&D.
He led the retrieval and ranking systems that brought the first generation of large language models into Baidu Search, and was among the early pioneers of this direction. Through systematic work in natural language understanding, his team achieved the #1 global ranking on the SuperGLUE Benchmark, earning Baidu's highest award.
He subsequently became a core contributor to the ERNIE series of large models, working across text understanding, multimodal understanding, and multimodal generation. The resulting work has been published at top venues including AAAI, CVPR, and KDD, accumulating over a thousand citations.
From the CNN-and-LSTM era, through BERT's rise, to the generative AI wave heralded by ChatGPT — Liu Jiaxiang has been a hands-on participant throughout this technological leap. Today, he is fully focused on the frontiers of generative AI, having developed ERNIE-Image, a text-to-image model that runs on consumer-grade GPUs.
Outside of work, he is drawn to photography — favoring spare, quiet imagery. He is also an avid and eclectic music listener.