Tootfinder

tschüss MFC-7820N und danke für viele, viele jahre treue dienste! willkommen MFC-L3760CDW, und auf viele gemeinsame bunte jahre! #heimbüro #bürotechnik #generationswechsel

Mind the Gap: A Closer Look at Tokenization for Multiple-Choice Question Answering with LLMs
Mario Sanz-Guerrero, Minh Duc Bui, Katharina von der Wense
https://arxiv.org/abs/2509.15020

Mind the Gap: A Closer Look at Tokenization for Multiple-Choice Question Answering with LLMs
When evaluating large language models (LLMs) with multiple-choice question answering (MCQA), it is common to end the prompt with the string "Answer:" to facilitate automated answer extraction via next-token probabilities. However, there is no consensus on how to tokenize the space following the colon, often overlooked as a trivial choice. In this paper, we uncover accuracy differences of up to 11% due to this (seemingly irrelevant) tokenization variation as well as reshuffled model rankings, ra…

Unlocking reversible and nonvolatile anomalous valley Hall control through multiferroic van der Waals heterostructures
Ankita Phutela, Saswata Bhattacharya
https://arxiv.org/abs/2508.13614

Unlocking reversible and nonvolatile anomalous valley Hall control through multiferroic van der Waals heterostructures
Achieving external control over the anomalous valley Hall (AVH) effect is essential for advancing valleytronic applications. However, many of the existing approaches suffer from limitations such as irreversibility or volatility. In this work, we propose a general strategy for enabling nonvolatile electrical tuning of the AVH effect by utilizing multiferroic van der Waals heterostructures. Using first-principles density functional theory calculations, we demonstrate that a heterostructure compos…

Source: Ravens' Van Noy bracing to miss time https://www.espn.com/nfl/story/_/id/46272073/source-ravens-kyle-van-noy-bracing-miss-multiple-games

Me and my new #yubikey5 part 2:
Now we get to the nitty-gritty parts. I'm using #mbsync to sync multiple #imap accounts to local

LEAP: LLM Inference on Scalable PIM-NoC Architecture with Balanced Dataflow and Fine-Grained Parallelism
Yimin Wang, Yue Jiet Chong, Xuanyao Fong
https://arxiv.org/abs/2509.14781

LEAP: LLM Inference on Scalable PIM-NoC Architecture with Balanced Dataflow and Fine-Grained Parallelism
Large language model (LLM) inference has been a prevalent demand in daily life and industries. The large tensor sizes and computing complexities in LLMs have brought challenges to memory, computing, and databus. This paper proposes a computation/memory/communication co-designed non-von Neumann accelerator by aggregating processing-in-memory (PIM) and computational network-on-chip (NoC), termed LEAP. The matrix multiplications in LLMs are assigned to PIM or NoC based on the data dynamicity to ma…

Kyle Van Noy injury: Ravens could be without star pass rusher for multiple games, per report

https://www.cbssports.com/nfl/news/kyle-van-no…

As usual we watched the "Stoertebekker" stage play at Ralswiek on Rügen Island. The stage setups and wardrobes were as awesome as always. They really try to improve on that every year and come up with clever mechanisms to transform the building for the various acts of the play. The script this year left room for desire ;) Nevertheless a happing worth attending
htt…

Types of graded tensor products of graded von Neumann algebras
Jumpei Tanaka
https://arxiv.org/abs/2508.10084 https://arxiv.org/pdf/2508.10084

Types of graded tensor products of graded von Neumann algebras
There is a famous multiplication table of types of tensor product of two von Neumann algebras. We filled out the multiplication table of graded tensor product of two graded von Neumann algebras in special cases.

OISMA: On-the-fly In-memory Stochastic Multiplication Architecture for Matrix-Multiplication Workloads
Shady Agwa, Yihan Pan, Georgios Papandroulidakis, Themis Prodromakis
https://arxiv.org/abs/2508.08822

OISMA: On-the-fly In-memory Stochastic Multiplication Architecture for Matrix-Multiplication Workloads
Artificial Intelligence models are currently driven by a significant up-scaling of their complexity, with massive matrix multiplication workloads representing the major computational bottleneck. In-memory computing architectures are proposed to avoid the Von Neumann bottleneck. However, both digital/binary-based and analogue in-memory computing architectures suffer from various limitations, which significantly degrade the performance and energy efficiency gains. This work proposes OISMA, a nove…