Safety pretraining has emerged as a promising paradigm for embedding safety deeply into a model during pretraining, yielding robustness to subsequent fine-tuning that conventional safety post-training often lacks. We refer to this safety acquired during model pretraining as native safety. Yet native safety is expensive...
Bingqi Shang, Yang Zhang, Yiwei Chen et al.· 0 citations
Retrospective Self-Distillation (RSD), a multi-round framework that turns such experience into dense supervision using privileged retrospective context, and frontier-guided context-policy evolution, where a stronger frontier model proposes constrained edits to how privileged context is constructed.
Yu-Hang Zhou, Yi-Fan Wu, Bo Peng et al.· 0 citations
Mixture-of-experts (MoE) models already make a sparse choice for every token: which experts should process it. We ask whether a sparse dictionary should reuse that choice. NR-SAE gives each routed expert a small dictionary and opens only those the frozen router selects. At block 15 of Gemma 4 26B-A4B, after ten million...
Large language models are deployed globally, yet safety alignment and evaluation remain concentrated in English: harmful requests refused in English can elicit unsafe completions when translated into low-resource languages. We evaluate ten widely deployed LLMs across 79 languages (English plus 78 low-resource) under a...
Jalil Alva, David Harper, Abigail E. Thornton et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
This work evaluates compaction on later questions of language-model agents, using four corpora that annotate the evidence of several questions per text and recorded agent trajectories, and compares every method with random deletion at the same budget.
Jun-Wei Yao, Yong-Xin Zhou, Zi-Ke Yang et al.· 0 citations
Effective detection of hate content is essential for maintaining healthy online communities, yet performance can vary substantially across languages. This work investigates a core practical question: Can training data from one language support hate-speech detection in another language with comparable performance? To an...
Preference labels align model training, but the judgments behind them are difficult to inspect or verify. In this paper, we examine how preference construction can be checked when intermediate decisions are recorded, and which of them remain uncertain. We present Symbolic Preference Labeling, a method in which a pretra...
Hallucination stress tests can yield similar average scores across language models even when different prompts elicit hallucinations in each model. We develop a procedure that iteratively edits source-grounded seed prompts to expose hallucinations in a target model. An LLM editor revises each prompt in response to the...
Jiho Lee, Bingquan Zhang, Lin Ting Liang et al.· 0 citations
Medical workflows often require multiple interdependent analyses, but existing specialist and generalist AI models are not designed to coordinate them. We introduce MMedAgent-Plus, a multimodal medical tool-use agent that combines a vision-language planner with 18 leading specialist models spanning 12 tasks and 8 modal...
Ruoxin Wang, Yuanting Pan, Binxu Li et al.· 0 citations
Profile-DP is introduced, a policy-aware local differential privacy framework designed specifically for user-profile release in personalized LLM inference, which provably bounds the release of user-withheld facts while treating public instructions, profile schemas, and authorized facts as public side information.
Mingchen Li, Zi-Fan Peng, Ke-Wei Sha et al.· 0 citations
Web agents are typically evaluated on English-language interfaces, leaving their ability to operate across the world's multilingual websites largely unexplored. This gap is especially significant for low-resource languages, where standardized resources for evaluating web agents remain limited. To address this gap, we i...
Ocean Monjur, Daud Ahmad Nisar, Shahriar Kabir Nahin et al.· 0 citations
Modern language models are increasingly asked to handle several tasks within a single prompt. Doing so requires solving each task correctly while integrating the resulting answers into a coherent response. In this paper, we identify a fundamental limitation in this process, where models make substantially more errors w...
Chenpan Li, Thies Brockmoeller, Stefanie Jegelka et al.· 0 citations
Flint-interactive gives AI agents a language for creating interactive visualizations. Click a chart, filter data, highlight related content, and explore beyond the view the author originally chose. The post Flint-interactive: A language for creating interactive data visualizations appeared first on Microsoft Research.
A new computational approach reveals how cells communicate to assemble into tissues. This could accelerate stem cell engineering, improve disease models, and inform therapies.
Microsoft Research Blog· microsoft.comOct 15, 2026
Zero-knowledge proofs can verify computing results without rerunning workloads or exposing sensitive data. Magnetar helps bring them into the mainstream with support for standard programming languages, major speed gains, and quantum-resistant security. The post Proof-based general-purpose computing with Magnetar appeared first on Microsoft Research.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
