致力于为硅基智能体倾力打造情感饱满、记忆超凡的智慧大脑
Creating emotional and intelligent brains for silicon-based agents
2026 AI Memory最新综述:从理论到实战,一文读懂AI记忆的进化全景 2026 AI Memory Survey: From Theory to Practice, Understanding the Evolution of AI Memory
快讯|北邮百家AI两篇工作被WWW 2026录用 Flash | Two papers from BUPT Baijia AI accepted by WWW 2026
智能体自主优化!LightSearcher工具调用减 39.6%、推理快 48.6% Agent Self-Optimization! LightSearcher reduces tool calls by 39.6% and speeds up inference by 48.6%
LLM工业级自进化:北邮与腾讯AI Lab提出MoE-CL架构,解决大模型持续学习核心痛点 Industrial LLM Self-Evolution: BUPT & Tencent AI Lab propose MoE-CL to solve Continual Learning bottlenecks
WSDM 2026| 句子级溯源大模型C²-Cite 重塑大模型可信度 WSDM 2026 | Sentence-level Attribution Model C²-Cite Reshapes LLM Trustworthiness
告别 "失忆" AI!首个大模型记忆操作系统(MemoryOS)开源框架来了! Say Goodbye to "Amnesia" AI! The first MemoryOS open-source framework for LLMs is here!
GMoE| 大模型优化密钥:GMoE 用图神经网络破解 MoE「冷热失衡」 GMoE | The Optimization Key: Solving MoE Load Imbalance via Graph Neural Networks
KG-Retriever | 北邮百家AI团队解锁垂域KG检索新范式 KG-Retriever | BUPT Baijia AI Team Unlocks New Paradigm for Domain Knowledge Graph Retrieval
北邮-腾讯联合提出多任务提示调优推荐框架MPT-Rec,实现效率性能双突破! BUPT & Tencent jointly propose MPT-Rec: Multi-Task Prompt Tuning for Recommender Systems