Microsoft Research Lab – Asia
新闻与深度文章
In this issue: Research Forum Ep. 4 explores multimodal AI. Registration is now open; Surveying developers’ AI needs; SuperBench improves cloud AI infrastructure reliability; Virtual Voices: Exploring factors influencing participation in virtual meetings.
Abstracts: July 29, 2024
| Gretchen Huizinga 和 Li Lyna Zhang
A lack of appropriate data, decreased model performance, and other obstacles have made it difficult to expand the input language models can receive. Li Lyna Zhang introduces LongRoPE, a method capable of extending content windows to more than 2 million…
The competitive dynamics of AI agents and a method for learning and applying temporal action abstractions represent just some of Microsoft’s contributions to ICML 2024.
Advancing time series analysis with multi-granularity guided diffusion model; An algorithm-system co-design for fast, scalable MoE inference; What makes a search metric successful in large-scale settings; learning to solve PDEs without simulated data.
Microsoft Research and Nissan Motor Corporation have collaborated to develop a machine learning model that improves the accuracy of predicting EV battery degradation by 80%. Learn how this collaboration supports long-term sustainability goals.
Unified databases offer better knowledge transfer between multimodal data types. They provide substantial corpus support for large language models and are poised to drive innovation in underlying hardware, laying the foundation for data-enhanced AI.
| Dongqi Han
The Bayesian behavior framework synergizes habits and goals through variational Bayesian methods, offering new insights on sensorimotor behavior and comprehension of actions.
Author: Chang Xu Diffusion probabilistic models have the capacity to generate high-fidelity samples for generative time series forecasting. However, they also present issues of instability due to their stochastic nature. In order to tackle this challenge, researchers from Microsoft Research…
In this issue: RELEVANCE automatically evaluates creative LLM responses; Recyclable vitrimer-based printed circuit boards; Lean Attention: Hardware-aware scalable attention mechanism; WaveCoder: a fine-tuned code LLM; New AutoGen training course.