Long-term memory is essential for large language model (LLM) agents operating in complex environments, yet existing memory designs are either task-specific and non-transferable, or task-agnostic but less effective due to low task-relevance and context explosion from raw memory retrieval. We propose PlugMem, a task-agnostic plugin…
This research looks at why Predictive Inverse Dynamics Models often outperform standard Behavior Cloning in imitation learning. By using simple predictions of what happens next, PIDMs reduce ambiguity and learn from far fewer demonstrations.
Microsoft AI (MAI) is seeking software engineers to help build the next wave of capabilities of our personal AI, Copilot. We’re looking for someone who possesses technical prowess, a methodical approach to problem-solving, proficiency in backend technologies and who will bring an abundance of positive…
Purpose To develop SCOPE (Small-lesion COntextual Pancreatic Evaluator), a deep learning model designed to improve CT detection of small pancreatic lesions—pancreatic ductal adenocarcinoma (PDAC), pancreatic neuroendocrine tumors (PanNETs), and cystic lesions—by integrating voxel-level features with global context. Materials and Methods This retrospective study used three independent datasets.…
Multi-turn jailbreaks capture the real threat model for safety-aligned chatbots, where single-turn attacks are merely a special case. Yet existing approaches break under exploration complexity and intent drift. We propose SEMA, a simple yet effective framework that trains a multi-turn attacker without relying on any…
At the intersection of public health and healthcare delivery, population health management analyzes population level data to identify opportunities to improve the quality, efficiency, and equity of care being provided. Population health management incorporates defining a population of individuals; recognizing the circumstances in which that…
This project studies the data protection challenges when models and agents interact with sensitive data across multiple users. Large language models (LLMs) are increasingly deployed in enterprise settings where thLarge language models (LLMs) are increasingly deployed in enterprise environments, where they serve multiple users and…
Microsoft Research unveils Paza, a human-centered speech pipeline, and PazaBench, the first leaderboard for low-resource languages. It covers 39 African languages and 52 models and is tested with communities in real settings.
Microsoft's Health and Life Sciences (HLS) group is driving a healthcare revolution with AI-powered solutions like Dragon Copilot. These innovations empower healthcare professionals to enhance patient care, improve outcomes, boost productivity, and reduce burnout. Our applied sciences team is looking for a Research Intern to…
AI has been increasingly integrated into screenwriting practice. In refinement, screenwriters expect AI to provide feedback that supports reflection across the internal perspective of characters and the external perspective of the overall story. However, existing AI tools cannot sufficiently coordinate the two perspectives to meet…
Recognizing what’s at stake, UNESCO (opens in new tab) designated 2022–2032 as the Decade of Indigenous Languages. This highlights a global effort to support revitalization and digital inclusion, and the work that partners in places like Nunavut are helping advance. According to the UNESCO Atlas of World’s Languages in…
Large Language Models (LLMs) are widely used for code generation. However, the correctness of code generated by LLMs remains a concern. A potential remedy to this concern is to have LLMs generate formal correctness proofs along with such code. However, compared with code generation, code-proof…
