Analyzing LLMs'internal processes can expose systematic vulnerabilities that warrant caution before deploying LLMs in strategic domains without robust guardrails, showing how analyzing LLMs'internal processes can expose systematic vulnerabilities that warrant caution before deploying LLMs in strategic domains without r...
Jan Sobotka, M. Karabag, U. Topcu· arXiv.org· 2 citations
Stories are key to transmitting values across cultures, but their interpretation varies across linguistic and cultural contexts. Thus, we introduce multilingual story moral generation as a novel culturally grounded evaluation task. Using a new dataset of human-written story morals collected across 14 language-culture p...
It is demonstrated that verbal confidence reflects automatic, sophisticated self-evaluation -- not post-hoc reconstruction -- with implications for understanding metacognition in LLMs and improving calibration.
D. Kumaran, Arthur Conmy, Federico Barbero et al.· arXiv.org· 21 citations· ⚡1
This work introduces MENASpeechBank, a reference speech bank comprising ~18K high-quality utterances from 124 speakers spanning multiple MENA countries, covering English, Modern Standard Arabic, and regional Arabic varieties, and develops a controllable data pipeline that constructs persona profiles enriched with WVS i...
Zien Sheikh Ali, Hunzalah Hassan Bhatti, R. N. Nandi et al.· arXiv.org· 4 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
It is found that the models'factual knowledge differs quite significantly from established knowledge bases, and that its accuracy is significantly lower than indicated by previous benchmarks, shedding light on future research opportunities in neuro-symbolic AI concerning extraction, consolidation and verification of fa...
Shrestha Ghosh, Luca Giordano, Yujia Hu et al.· 2 citations
A novel method is introduced that localizes the model's few-shot learning ability to only a few attention heads through dimensionality reduction and decomposition of the heads'output spaces, and derives a mathematical identity relating the''aggregator''and''extractor''subspaces of attention heads, allowing us to track...
Xin-Yang Hu, Kayo Yin, Michael I Jordan et al.· arXiv.org· 12 citations· ⚡2
It is suggested that robust meme understanding requires multimodal training, varies substantially across semantic categories, and remains sensitive to over-specialization when models are fine-tuned on individual datasets rather than trained in a unified setting.
Ali Ezzat Shahroor, Mohamed Bayan Kmainasi, A. Hasnat et al.· arXiv.org· 5 citations
DiaVLo is presented, a diagnostic framework that leverages human curation and VLMs'generation capabilities to construct specifications of desired and observed VLM behaviours, surfacing potential misalignments and providing causal estimates to identify the most influential concepts steering VLM behaviours.
The results demonstrate that full-duplex interaction, speech recognition and generation, general language capabilities, and external tool use can be integrated in a single open speech-to-speech model without sacrificing real-time conversational behavior.
Jagadeesh Balam, Travis M. Bartley, Edresson Casanova et al.· 2 citations· ⚡1
This work introduces an inclined boundary that evaluates prediction loss relative to predictive entropy, and shows that entropy correction can preserve the expected membership signal while reducing its variance, thereby improving standardized member--non-member separation.
Chen-Ye Ke, Zi-Rui Liu, Qi Liu et al.· 0 citations
This work investigates whether personality-aware fine-tuning can reduce the consistency and controllability of personality-conditioned dialogue generation compared with instruction prompting alone, and indicates that fine-tuned models are not better at role-playing different personalities than their respective baseline...
The results show that LLMs continue to struggle with the cross-lingual understanding of Chinese internet buzzwords, particularly in fine-grained non-literal interpretation, robust equivalent matching under option perturbations, and calibrated harmfulness detection, which highlights the persistent challenges posed by cu...
Yi-Fan Wang, Jun-Yu Lu, Qi-Fan Wang 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.
