🎓How I Study AIHISA

📖Read

📄Papers 📰Blogs 🎬Courses

💡Learn

🛤️Paths 📚Topics 💡Concepts 🎴Shorts

🎯Practice

🧩Problems 🎯Prompts 🧠Review

How I Study AI - Learn AI Papers & Lectures the Easy Way

Papers2

All Beginner Intermediate Advanced

All Sources arXiv

#KV Cache

DiRL: An Efficient Post-Training Framework for Diffusion Language Models

Ying Zhu, Jiaxin Wan et al.Dec 23arXiv

This paper builds DiRL, a fast and careful way to finish training diffusion language models so they reason better.

#Diffusion Language Model#Blockwise dLLM#Post-Training

SWAA: Sliding Window Attention Adaptation for Efficient Long-Context LLMs Without Pretraining

Yijiong Yu, Jiale Liu et al.Dec 11arXiv

Long texts make standard attention in large language models very slow because it checks every word against every other word.

#Sliding Window Attention#SWAA#FA Decode