A systematic, controlled study of the interplay between training language coverage, model scale, and task domain, based on 220 supervised fine-tuning runs on parallel translated multilingual data mixtures spanning mathematical reasoning and API calling tasks, finds that English-only post-training is typically suboptima...
Mehak Preet Dhaliwal, S. Chaurasia, Yao Qin et al.· arXiv.org· 0 citations
It is found that calibration shifts substantially after SFT: across 216 configurations, it degrades in 112 cases and improves in 104, with confidence scores shifting due to factors beyond output quality, such as proximity to the training distribution.
Lorenzo Jaime Yu Flores, Cesare Spinoso di-Piano, Jackie C. K. Cheung· 0 citations
MB using only 10% of the target-domain speech achieves word error rates comparable to, or better than, conventional ASR fine-tuning with the full dataset, indicating that small amounts of speech provide a strong modality-alignment signal.
Large Reasoning Models (LRMs) still exhibit large performance gaps between English and other languages, yet much current work assumes these gaps can be closed simply by making reasoning in every language resemble English reasoning. This work challenges this assumption by asking instead: what actually characterizes succ...
Dayeon Ki, Kevin Duh, Marine Carpuat· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Outliers in dynamic topic modeling are typically treated as noise, yet we show that some can serve as early signals of emerging topics. We introduce a temporal taxonomy of news-document trajectories that defines how documents relate to topic formation over time. It distinguishes anticipatory outliers, which precede the...
Evangelia Zve, Gauvain Bourgne, Benjamin Icard et al.· 0 citations
This paper considers multilingual pretraining as a cooperative game in which each language acts as a player that jointly contributes to pretraining, gaining the resulting reduction in test loss as the payoff and proposes a game-theoretic multilingual scaling law called ShapleyLaw.
Associations between speech sounds and meaning are well documented but have not been systematically mapped over a whole alphabet. Here we map them across the 26 English letters and find that each carries a structured, multidimensional semantic profile that is recoverable from text, perceived across languages, and predi...
CCTU, a benchmark for evaluating LLM tool use under complex constraints, is introduced and nine state-of-the-art LLMs are evaluated in both thinking and non-thinking modes, showing that models violate constraints in over 50% of cases.
Junjie Ye, Guoqiang Zhang, Wen-Jie Fu et al.· arXiv.org· 8 citations
This work introduces MEDGPT-OSS, an open-weight, 20B-parameter generalist vision-language model designed to facilitate open research in clinical AI, which successfully outperforms larger open medical models on out-of-distribution multimodal reasoning and complex text-only clinical tasks.
Kai Zhang, Zheng-Qing Yuan, Cheng Peng et al.· arXiv.org· 0 citations
SVF is proposed, which learns a differentiable concept scoring function whose local gradient defines the steering direction at each activation, making interventions explicitly context-dependent and enables efficient long-form and multi-attribute control within a unified framework.
The global deployment of large language models (LLMs) has raised concerns about cultural misalignment, yet the linguistic properties of fine-tuning datasets used for cultural adaptation remain poorly understood. We adopt a dataset-centric view of cultural alignment and investigate which properties of post-training data...
Reem I. Masoud, Chen Feng, Shunta Asano et al.· 0 citations
It is shown that attention heads with high BOS sink scores are strongly associated with functional redundancy: such heads, especially in deeper layers, contribute little to predictive performance and effectively serve as dumping grounds for superfluous attention weights.
Jaewon Sok, J. Yeom, Seonghyeon Park et al.· arXiv.org· 8 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.
