Deception probes on language models answer a binary question: is this response deceptive? Charts allow a graded one, because a chart rendered from its own distortion parameters has an exact Tufte Lie Factor. We render 1,685 charts from real statistical series under parameterised distortions and train linear probes on t...
Carson CONCEPTION Rodrigues· Zenodo (CERN European Organi...· 0 citations
TDLT_YasudaK π Ab Initio Transcendence What is presented here constitutes only a small sample of several years of independent research. This work does not ask the reader to accept The Dark Light Theory (TDLT) through persuasion. TDLT is proposed as a continuous dependency architecture, in which later structures are int...
SILVIO KOZO YASUDA· Zenodo (CERN European Organi...· 0 citations
Short takeaway: Mnemorphics is a mega‑document—a merged research codex—combining acoustic metamaterials, shockwave diodes, energy‑harvesting physics, AI cognitive‑warfare systems, particle‑capture engineering, and large‑scale defensive architecture. It reads like a fusion of hard engineering, corrected physical models,...
francis lee· Zenodo (CERN European Organi...· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Parameter-efficient fine-tuning is widely assumed to improve small open-weight models on the specific behaviors it targets. We present a controlled case study that complicates this assumption for multi-tool agent execution. Using Qwen2.5-7B-Instruct and a fixed 23-task, machine-checkable benchmark, we show that an init...
Alyssa L· Zenodo (CERN European Organi...· 0 citations
What a buyer abroad already believes before a seller says anything A buyer three time zones away decides about a supplier before the supplier has said anything, on evidence the supplier did not know was being read. A small manufacturer trying to sell abroad believes the problem is reach, and puts its effort into approa...
Alya Iskandar· Zenodo (CERN European Organi...· 0 citations
It is demonstrated that task-specific fine-tuning enables open-source LLMs to rival proprietary LLMs and expert radiologists in Chinese ultrasound report error detection, offering a locally deployable and privacy-preserving, and regulation-compliant solution for enhancing reporting consistency and reducing diagnostic e...
Min Lai, Jin Zhang, Ya-Er Lv et al.· Insights into Imaging· 0 citations
TDLT_YasudaK π Ab Initio Transcendence What is presented here constitutes only a small sample of several years of independent research. This work does not ask the reader to accept The Dark Light Theory (TDLT) through persuasion. TDLT is proposed as a continuous dependency architecture, in which later structures are int...
SILVIO KOZO YASUDA· Zenodo (CERN European Organi...· 0 citations
What a buyer abroad already believes before a seller says anything A buyer three time zones away decides about a supplier before the supplier has said anything, on evidence the supplier did not know was being read. A small manufacturer trying to sell abroad believes the problem is reach, and puts its effort into approa...
Alya Iskandar· Zenodo (CERN European Organi...· 0 citations
Inverted perovskite solar cells (PSCs) possess great potential for improving efficiency and operating stability, but parasitic charge recombination at the interfaces still limits their performance. As additives and surface treatment agents, ammonium ligands have been reported to form either two/three-dimensional (2D/3D...
Yu-Lan Huang, Mu-Chen Li, Qing-Bin Cai et al.· Advances in Materials· 0 citations
Parameter-efficient fine-tuning is widely assumed to improve small open-weight models on the specific behaviors it targets. We present a controlled case study that complicates this assumption for multi-tool agent execution. Using Qwen2.5-7B-Instruct and a fixed 23-task, machine-checkable benchmark, we show that an init...
Alyssa L· Zenodo (CERN European Organi...· 0 citations
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.
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
