DeepSeek’s conditional memory fixes silent LLM waste: GPU cycles lost to static lookups

Through systematic experiments DeepSeek found the optimal balance between computation and memory with 75% of sparse model ...
GitHub

Dynamic Cheatsheet: Test-Time Learning with Adaptive Memory

A lightweight framework that gives language models (LMs) a persistent, evolving memory during inference time. Dynamic Cheatsheet (DC) endows black-box language models with the ability to store and ...
IEEE

Static and Dynamic Routing, Fiber, Modulation Format, and Spectrum Allocation in Hybrid ULL Fiber-SSMF Elastic Optical Networks

Abstract: Traditional standard single-mode fibers (SSMFs) are unable to satisfy the future long-distance and high-speed optical channel transmission requirement due to their relatively large signal ...
IEEE

Flight Attendant Route Planning and Resource Allocation Algorithm Based on Reinforcement Learning and Dynamic Programming

Abstract: In this paper, flight attendants' route planning and resource allocation algorithms based on feedback learning (RL) and dynamic programming (DP) are studied. In the air transport industry, ...