Xnhyacinth/Awesome-LLM-Long-Context-Modeling

What it solves

Keeping up with the fast-moving field of long-context language modeling is difficult because research spans many subtopics and releases continuously. This project provides a single, organized place to find relevant papers and blog posts. It helps researchers and practitioners avoid missing important work and quickly see what is happening in areas like efficient attention, memory management, and evaluation.

How it works

It is a curated "Awesome" list: a structured collection of links to papers and blogs, grouped into categories such as efficient attention, KV-cache optimization, memory, compression, reasoning, video, agents, and evaluation. The list is updated weekly and monthly with the latest arXiv papers, many including links to code or project pages. It also includes a survey paper (arXiv:2503.17407) and a visual taxonomy diagram to help navigate the field, making it easy to drill into a specific subtopic.

Who it's for

Researchers, students, and practitioners who work on or are interested in long-context language models. If you need a comprehensive, up-to-date overview of the state of the art without having to search across many sources, this repository saves time and keeps you informed.

Highlights

• Covers 20+ distinct topics from efficient attention and KV-cache optimization to long-horizon agents and benchmarks
• Weekly and monthly updates keep the list current with the newest research
• Includes a comprehensive survey paper (arXiv:2503.17407) and collaboration with LCLM-Horizon
• Community-driven with contributions welcome; has many stars, forks, and contributors
• Provides a "Taxonomy at a glance" diagram for quick orientation