October 11, 2023

编者按：欢迎阅读“科研上新”栏目！“科研上新”汇聚了微软亚洲研究院最新的创新成果与科研动态。在这里，你可以快速浏览研究院的亮点资讯，保持对前沿领域的敏锐嗅觉，同时也能找到先进实用的开源工具。 本期内容速览 01. AniPortraitGAN：可驱动的真实感3D肖像生成 02. KOSMOS-2.5：阅读文本密集图像的多模态大型语言模型 03. PromptTTS 2：利用文本描述创造语音合成的音...

October 9, 2023 | Gretchen Huizinga and Sheng Zhang

Researcher Dr. Sheng Zhang joins “Abstracts”—your source for cutting-edge research in brief—to discuss a recent paper on distilling large language models into smaller, more efficient ones capable of excelling in broad application classes.

October 6, 2023 | Li Lyna Zhang, Jiahang Xu, Quanlu Zhang, Yuqing Yang, Ting Cao, and Mao Yang

A persistent challenge in deep learning is optimizing neural network models for diverse hardware configurations, balancing performance and low latency. Learn how SpaceEvo automates hardware-aware neural architecture search to fine-tune DNN models for swift execution on diverse devices.

October 5, 2023 | Xin Wang and Neel Joshi

HoloAssist is a new multimodal dataset consisting of 166 hours of interactive task executions with 222 participants. Discover how it offers invaluable data to advance the capabilities of next-gen AI copilots for real-world tasks.

October 5, 2023 | Madeleine Daepp, Alejandro Cuevas, and Jennifer Scurrell

Connecting with researchers, collaborating across disciplines, and exploring a new city—PhD students Jennifer Scurrell and Alejandro Cuevas talk to Senior Researcher Madeleine Daepp about the internship experience at Microsoft Research.

October 4, 2023

Ronen Eldan (Microsoft Research) and Mark Russinovich (Azure) The Challenge of Unlearning in an AI Era  Over the last few months, significant public attention has focused on a wide variety of questions related to the data used to train large…

October 4, 2023

Ronen Eldan and Mark Russinovich The Challenge of Unlearning in an AI Era  Over the last few months, significant public attention has focused on a wide variety of questions related to the data used to train large language models (LLMs).…