Why is SIG Public Affairs important for the GÉANT community right now?
Between evolving digital infrastructure strategies, AI regulation and complex funding cycles, the rules of the game are moving fast, often with NRENs navigating it alone.
That's exactly the gap this newly established SIG was created to fill.
To learn more, we recently spoke to Agnieszka Stokłosa (PCSS) & Leonie Schäfer (DFN), Chairs and Strategic Coordinators of the SIG.
🔗 Read the inte…
5 Coaches Who Could Take the Cowboys to the Next Level https://insidethestar.com/5-coaches-who-could-take-the-cowboys-to-the-next-level
The Defense Department’s announcement that it would
live-stream the scheduled execution of Fort Hood shooter Nidal Malik Hasan
reverses decades of established practice surrounding the death penalty.
If carried out, it would be the first known time that U.S. authorities have broadcast live video of an execution to the public.
It would also be the first public execution in the U.S. in decades,
the first U.S. military execution since 1961,
and the first mil…
#WordWeavers 10/9: Share a description from your WIP that establishes atmosphere without explicitly naming the emotion.
This is a passage from my upcoming book, “The Management Apologises for the Inconvenience”.
Welcome to #TheMentalMansion
Global warming will exceed limit, UN says in a report that maps path to get back below danger zone https://phys.org/news/2026-09-global-exceed-limit-path-danger.html
UN report finds 1.5°C warming limit likely to be exceeded by 2030
APEX: Adaptive Expert Prefetching for Memory-Efficient Edge MoE Inference
Alish Kanani, Layan Badawi, Umit Y. Ogras
https://arxiv.org/abs/2608.11688 https://arxiv.org/pdf/2608.11688 https://arxiv.org/html/2608.11688
arXiv:2608.11688v1 Announce Type: new
Abstract: Mixture-of-Experts (MoE) models are attractive for edge deployment because they provide high model capacity while activating only a small subset of parameters per token, improving compute efficiency. However, MoE inference at the edge is fundamentally limited by memory. Expert parameters are large and often reside in off-chip memory due to capacity, cost, and power constraints, putting expert loading to the critical path. We present APEX: Adaptive Expert Prefetching, a predictive resource management framework that overlaps expert loading with useful computation. APEX introduces a lightweight prefetch router that predicts candidate experts before the attention block to dynamically fetch additional experts using a learned confidence model. This adaptive strategy achieves over 99% overlap accuracy, significantly outperforming fixed top-k prefetching techniques. APEX supports two execution modes: a correctness-preserving mode that guarantees exact routing semantics, and a stall-free mode that eliminates residual stalls by operating on available experts with negligible impact on application accuracy. Across multiple MoE models, the correctness-preserving mode reduces per-token latency by up to 26% and improves energy-delay product (EDP) by up to 41% over state-of-the-art baselines, while the stall-free mode provides additional efficiency gains with negligible impact on application accuracy. These results establish adaptive, confidence-driven expert prefetching as an effective approach for efficient MoE inference on edge systems.
toXiv_bot_toot
Explicit Estimates for the Bergman Kernel Form on Polarized Riemann Surfaces for General Tensor Powers
Johannes Testorf
https://arxiv.org/abs/2608.03493 https://arxiv.org/pdf/2608.03493 https://arxiv.org/html/2608.03493
arXiv:2608.03493v1 Announce Type: new
Abstract: Let $(L,e^{-\phi})$ be a positive Hermitian holomorphic line bundle over a compact Riemann surface $X$, and put $\omega=\ddbar\phi$. We obtain effective pointwise estimates for the Bergman form of $H^0(X,K_X\otimes L^m)$. If $\Ric\omega\leq\omega$ and the shortest nonconstant closed geodesic has length at least $2\pi$, then \[
K_{m\phi}\geq \frac{2m-1}{4\pi}\,\omega, \] and the constant is sharp on $(\mathbb P^1,\mathcal O_{\mathbb P^1}(2))$. A local version, depending on an upper curvature bound and the injectivity radius, recovers the first two terms of the Bergman expansion when the curvature is constant. Under the two-sided bound $-\omega\leq\Ric\omega\leq\omega$ and the same closed-geodesic hypothesis, we also prove \[ K_{m\phi}\leq \frac{m\omega}{2\pi}
\left(1 \frac{54.8\log(2m)}{m-\frac{1}2}\right). \] The lower estimates use the deformation-to-the-tangent-space form of the Ohsawa--Takegoshi theorem established by He, Wang, and the author, whereas the upper bound combines a weighted submean inequality with quantitative isothermal coordinates which was obtained in recent work by Eilat.
toXiv_bot_toot
Spain proposes 23 new marine protected areas in the Canary Islands and Mediterranean https://news.mongabay.com/2026/08/spain-proposes-23-new-marine-protected-areas-in-the-canary-islands-and-mediterranean/