59 citations · 124 across the 11 of their papers we have counts for
11 papers
Joint Encoding of KV-Cache Blocks for Scalable LLM Serving
Joseph Kampeas, Emir Haleva
Modern large language models (LLMs) drive interactive AI systems but are bottlenecked by the memory-heavy growth of key-value (KV) caches, which limits real-time throughput under c…
Linear Log-Normal Attention with Unbiased Concentration
Yury Nahshan, Joseph Kampeas, Emir Haleva
Transformer models have achieved remarkable results in a wide range of applications. However, their scalability is hampered by the quadratic time and memory complexity of the self-…
Rotation Invariant Quantization for Model Compression
Joseph Kampeas, Yury Nahshan, Hanoch Kremer +4
Post-training Neural Network (NN) model compression is an attractive approach for deploying large, memory-consuming models on devices with limited memory resources. In this study,…
Vesicles as osmotically stressed capsules
Emir Haleva, Haim Diamant
Vesicular capsules are used to carry biochemicals in biology and liposome technology. Being water-permeable with differing interior and exterior compositions, they are necessarily…
Permeability of phospholipid membrane for small polar molecules determined from osmotic swelling of giant phospholipid vesicles
Primoz Peterlin, Vesna Arrigler, Haim Diamant +1
A method for determining permeability of phospholipid bilayer based on the osmotic swelling of micrometer-sized giant unilamellar vesicles (GUVs) is presented as an alternative to…
Law of corresponding states for osmotic swelling of vesicles
Primoz Peterlin, Vesna Arrigler, Emir Haleva +1
As solute molecules permeate into a vesicle due to a concentration difference across its membrane, the vesicle swells through osmosis. The swelling can be divided into two stages:…