Anthropic Made Claude Forget Most of a Task. Sometimes It Got Better.
中文摘要
Anthropic的研究显示，让Claude遗忘部分任务细节有时反而能提升性能，这解释了为何长对话效果会下降。
English Summary
Anthropic's research shows that making Claude forget certain task details can sometimes improve performance, explaining why long AI conversations often degrade.
Original Excerpt
A new robotics study reveals why longer AI conversations can feel strangely worse — and why useful memory starts with knowing what to… Continue reading on Medium »