The Agent Architect · 2026-W32
The Agent Architect #32: Transactive Memory Systems
收听最新一期 · 1 min
本周模式
Transactive Memory Systems
- 是什么:
- 各智能体共同维护彼此专长的认知，并按专业分工分派任务，从而扩展集体解决问题的能力。
- 何时使用:
- 多智能体团队处理复杂项目，各智能体技能各异，需要高效分派工作而避免重复处理。
- 注意:
- 专长图谱很快就会过时；智能体可能把任务派给已不再擅长的一方，或忽略能力变化，导致性能随时间下降。
本周智能体AI动态
- How GPT-5.6 fuses frontier intelligence with frontier efficiencyOpenAI
GPT-5.6 improves efficiency across models, inference, and agentic workflows.
- LFM2.5-Encoders for Fast Long-Context Inference on CPUHugging Face
LFM2.5-Encoders provide fast long-context inference on CPU hardware.
- How enabling two settings tripled our scores on the ARC-AGI-3 benchmarkOpenAI
Two API settings improved GPT-5.6 performance on ARC-AGI-3 by retaining reasoning and enabling compaction.
- Grounding Agentic VLMs with Dedicated Segmentation for Fine-Grained Vehicle Damage AssessmentarXiv cs.AI
Dedicated segmentation improves vision-language model spatial grounding for fine-grained vehicle damage assessment tasks.
- Cultural Awareness is Represented but Not Decoded: Tracing Mythological Knowledge across 18 Open-Source LLMsarXiv cs.AI
Open-source LLMs encode Western mythology more reliably than non-Western traditions; cultural bias originates in specific model layers.
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。