直接挑战 LLM memory consolidation 的常见设计,给 Hermes 记忆系统提供了反证。
About
AI Agent & Full-Stack Developer
Undergraduate at the University of Melbourne. Currently building atypica (a multi-agent system for commercial research) at Tezign, and working on AI full-stack features and a video-editing agent at fAIshion.ai and Goshu Tech. I care about how LLMs actually work under the hood, and about shipping products that run in the real world. Build fast, learn faster — off the clock: piano, cello, photography.
More about me
Skills
Languages
Frontend
Backend
AI & Agent
Tools
Curated
- Paper UIUC / TsinghuaUseful Memories Become Faulty When Continuously Updated by LLMs — Zhang et al.
- Repo github.comHermes Agent — hermes_state.py + session_search_tool.py
单用户 agent 场景下,搜索系统可以先做词法召回,再把理解交给 LLM。
More curated
Site stats
12
Posts
11
Notes
12
Projects
5,215
Page views