The proposed alignment method is statistically consistent and yields significantly tighter convergence guarantees than DDRO, and is based on the relative density ratio between the preferred data distribution and a mixture of the preferred and non-preferred data distributions.
This first rigorous pipeline-wide study of structural bias across major phases of LLM development shows that contemporary LLMs privilege AmE as the de facto norm, raising concerns about linguistic homogenization, epistemic injustice, and inequity in global AI deployment, while providing a rigorous basis for targeted co...
This work presents an inspectable LLM Council with two synthesis stages: an analyst organizes three independent candidate answers into a structured textual state, and a separately invoked writer composes the final response from the question and that state.
Standard accuracy flatters all ten models the authors evaluate and reorders their ranking, and prompting fixes that look effective largely vanish under paired scoring.
Yubo Li, Lu Zhang, Tian-Chong Jiang et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
The ability to represent oneself and others as agents with knowledge, intentions, and belief states that guide their behavior - Theory of Mind - is a human universal that enables us to navigate - and manipulate - the social world. It is supported by our ability to form mental models of ourselves and others. Its ubiquit...
Large language models (LLMs) frequently exhibit performance biases against regional dialects of low-resource languages. However, frameworks to quantify these disparities remain scarce. We propose a two-phase framework to evaluate dialectal bias, operationalized as comprehension degradation relative to standard Bengali,...
K. M. Jubair Sami, Dipto Sumit, Ariyan Hossain et al.· 0 citations
A three-stage evaluation pipeline is designed that progressively reduces textual guidance to systematically evaluate LALMs' in-context learning ability in the audio modality, uncovering a consistent asymmetry across LALMs.
Yen-Ting Piao, Jay Chiehen Liao, W. Chien et al.· 1 citation
IndexRAG is presented, a novel approach that shifts cross-document reasoning from online inference to offline indexing and achieves the best average performance among all evaluated methods, including graph-based baselines such as HippoRAG2 and FastGraphRAG, while relying on a flat vector index.
Sampling from a categorical distribution is mathematically simple, but in large-vocabulary decoding, it often triggers extra memory traffic and extra kernels after the LM head. We present FlashSampling, an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in H...
Tomas Ruiz, Zhen Qin, Yifan Zhang et al.· 0 citations
SalamahBench is introduced, a unified benchmark of 8{,}270 human-verified harmful prompts across ML Commons hazard categories, each rendered in Modern Standard Arabic (MSA) and five regional Arabic varieties and five regional Arabic varieties, for a total of 49{,}620 paired instances.
Omar Abdelnasser, Fatemah Alharbi, Khaled N. Khasawneh et al.· 0 citations
This study investigates the diverse characteristics of nouns, focusing on both semantic and morphosyntactic distinctions, and proves the model's representational completeness in learning and prediction and presents a unified framework for understanding the complexities of morphological markings across languages.
This study provides new mechanistic insights into how LLMs internalize OOD challenges, and designs Sparsity-Guided Curriculum In-Context Learning (SG-ICL), a strategy that explicitly uses representation sparsity to schedule few-shot demonstrations, leading to considerable performance enhancements.
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.
MIT News · Artificial Intelligence· news.mit.eduSep 24, 2026
