InfoQ

MIT's Recursive Language Models Improve Performance on Long-Context Tasks

Researchers at MIT's CSAIL published a design for Recursive Language Models (RLM), a technique for improving LLM performance ...
IEEE

Trend and Order Features for Semi-Supervised Time-Series Classification via Multitask Learning

Abstract: Multitask learning with a pretext task has excelled in time-series classification task lacking labeled data. The key to multitask learning is to build a pretext task and learn the most ...
IEEE

Self-Supervised Learning for Spaceborne SAR and Multispectral Image Representation With Few-Shot Local Climate Zone Classification

Abstract: Deep learning has demonstrated significant potential in remote sensing scene classification; nevertheless, its efficacy is frequently hindered by the requirement for extensive labeled ...