Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but...
Jaewon Sok, Jewon Yeom, Seonghyeon Park et al.· 0 citations
Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a learned ability to propose effective improvements across models faces two challenges. First, t...
Min-hua Lin, Zi-Jun Wang, Ze-Wen Liu et al.· 0 citations
This work investigates a token probability approach that scores each text sample as a continuation of a prompt prefix specifying a candidate L1, then predicts the candidate with the highest score, which substantially improves on direct classification.
Ahmet Yavuz Uluslu, M. Gales, Kate Knill et al.· 0 citations
VERITAS is introduced to verbalise the model's computation: how different reasoning anchors emerge, transfer across layers, and surface in the output, to provide a richer and more interpretable lens on CoT faithfulness.
Yu-Jing Wang, Yu-Lan He, Wen-Ya Wang et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Tool-using agents read untrusted external content, which exposes them to indirect prompt injection (IPI), where hidden instructions hijack the agent's tool calls. Benchmarks measure IPI security by the attack success rate (ASR), which treats every failed attack as a successful defense. To test this assumption, we audit...
Farhan Zarif, Kazi Shahid Ahmed Galib, Synthia Akter Promi et al.· 1 citation
A systematic evaluation of confidence-based routing across three model families, four languages, and three task types, covering 72 model, dataset, language, and scale configurations and 42,288 predictions shows that confidence routing is task-dependent and can allocate computation unevenly across languages.
GRPO assigns the same group-relative advantage to equally rewarded responses within a question, leaving their generation uncertainty unused. We introduce MGE-PO, centered on four-quadrant (4Q) advantage shaping that combines verified correctness with rollout generation entropy. Correctness determines the sign of an add...
Ji-Hao Huang, Guang-Jia Chai, Zhi-Rui Chen et al.· 0 citations
In a hierarchical multi-agent system, a Leader assigns tasks and subagents carry them out. Jointly training the Leader and the subagents is difficult because a final answer gives only indirect feedback to the subagents. An LLM judge can score the performance of the subagents, but these scores depend on the judge's own...
Yuhan Sun, Jiaxin Zhao, Yongdong Luo et al.· 0 citations
Drifting Models generate a sample with a single function evaluation, making them attractive for fast parallel language generation, but their one-step formulation does not support iterative refinement. We propose Recurrent Language Generation via Drifting (ReLaDrift), which uses a shared Drifting Model to generate an in...
Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-t...
Jingting Zheng (TJUNLP Lab, School of Computer Science and Technology, Tianjin University et al.· 0 citations
Text detoxification requires balancing toxicity removal, semantic preservation, and fluency. Existing methods typically optimize an unconstrained scalar objective---enabling safety scores to mask semantic loss---or rely on open-ended iterative debate prone to semantic drift and high computational cost. We propose AMEND...
Detecting faithfulness hallucinations in large language models (LLMs) is crucial for their safe deployment in high-stakes domains. Recent approaches address this by contrasting contextual against parametric knowledge utilization, yet measuring context reliance typically demands noticeable overhead, such as materializin...
A. Kuleshova, Andrey Volodichev, Daria Kotova 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.
