yyyujintang / Awesome-Agent-Memory-Papers

What is Awesome-Agent-Memory-Papers?

This repo curates 90 papers on memory systems for LLM and multimodal agents, covering methods, benchmarks, and surveys across episodic, semantic, procedural, and multimodal types—both internal parametric storage and external retrieval setups. It organizes everything into clear categories like QA benchmarks, web navigation evals, and GUI tasks, with an interactive dashboard on GitHub Pages for multi-tag filtering. Like other awesome agent papers GitHub lists or awesome AI papers, it saves you hours hunting arXiv for agent memory research.

Why is it gaining traction?

Its tight categorization and tag legend (storage type, learning method, memory flavor) beat scattered awesome LLM papers or awesome RAG papers collections, plus the live-updating dashboard makes browsing benchmarks for web agents or embodied envs dead simple. Recent papers up to 2026 keep it fresh amid the agent boom, and open PRs invite community fixes—unlike static awesome CV papers repos.

Who should use this?

AI researchers prototyping long-horizon agents for web nav or desktop GUIs, benchmark hunters evaluating memory in production LLM setups, or devs tuning multimodal agents with episodic recall. Skip if you're deep in awesome GitHub Copilot prompts; this targets agent memory specifically.

Verdict

Solid starting point for agent memory lit review despite low 1.0% credibility score and 44 stars signaling early maturity—docs are crisp, but expect gaps in older papers. Star it if you're in awesome agents papers territory; fork for your own awesome machine learning papers spin. (187 words)