New KV cache compaction technique cuts LLM memory 50x without accuracy loss

MIT researchers developed Attention Matching, a KV cache compaction technique that compresses LLM memory by 50x in seconds — ...

Medicus Pharma Ltd: Medicus Pharma Reports Positive Phase 2 SKNJCT-003 Topline Data Observing 73% Clinical Clearance and 40% Histological Clearance (CR) at Day 57 in 200µg Cohort

Medicus Pharma Ltd. (NASDAQ: MDCX) ("Medicus" or the "Company"), a biotech/life sciences company focused on advancing the clinical development programs ...
EurekAlert!

Real-time multi-objective charging scheduling unlocks efficient renewable integration and grid-friendly electric vehicle operations at public stations

The global surge in electric vehicle adoption stands as a cornerstone of efforts to combat climate change and reduce reliance ...
American Institute of Physics

Multi-objective algorithm speeds up UAV search-and-rescue

Incorporating multiple constraints such as task completion time, UAV payload capacity, and flight range into path optimization algorithms allows for more efficient search patterns.

Qwen3.5 family: Fireworks of new LLMs from Alibaba

In the last few days, Qwen then set off real fireworks with new models. Qwen started with the large models Qwen3.5-122B-A10B, ...
Treasure Coast Newspapers

Florida Physician Advances Multi-Specialty Regenerative Standard at MIT Florida

Genesis Regenerative emphasizes patient outcomes and clinical integrity over profits, which is why I chose to lead these initiatives.” — Mark D. Jacobson, MD, Medical Director of Genesis Regenerative ...

Aptevo Therapeutics: Mipletamig Delivers Compelling 86% Clinical Benefit Rate and No CRS as Evaluable AML Patient Data increases by Nearly 50%

Frontline patient outcomes are making the case for mipletamig to enhance standard-of-care therapy alongside venetoclax + azacitidine SEATTLE, WA / ACCESS Newswire / March 10, 2026 / Aptevo Therapeutic ...
eLife

Designing optimal perturbation inputs for system identification in neuroscience

To enable more accurate estimation of connectivity, we propose a data-driven and theoretically grounded framework for optimally designing perturbation inputs, based on formulating the neural model as ...