1
Compaction amnesia is why you keep re-explaining
大模型会话压缩后失忆，导致你反复解释？破解长对话衰减的秘诀在这。
TL;DR: Compaction is lossy by design, and the loss compounds across passes. After a few compactions, your agent remembers the gist but loses the why —…
2
DYCP: Dynamic Context Pruning for Long-Form Dialogue with LLMs
长对话中动态修剪上下文，大幅提升LLM效率与性能，看完直呼“省省你的token”！
arXiv:2601.07994v5 Announce Type: replace Abstract: Large Language Models (LLMs) increasingly operate over long-form dialogues with frequent topic shi…
3
OpenAI 调整手机端 ChatGPT，提示词可提前选 AI 响应档位
手机端ChatGPT新增隐藏手势，发提示词前可提前选AI“智能程度”，对话体验更灵活。
IT之家 6 月 2 日消息，科技媒体 9to5Mac 昨日（6 月 1 日）发布博文，报道称 OpenAI 升级其 ChatGPT 应用，聚焦更新提示词发送、长对话浏览和默认模型体验上。 在手势交互方面，移动端新增隐藏手势，用户长按发送箭头，就能在发出提示词（Prompt）前选择“智能程度”。IT…
4
Cooperative Memory Paging with Keyword Bookmarks for Long-Horizon LLM Conversations
利用关键词书签实现协作内存分页，突破LLM长对话上下文窗口限制
arXiv:2604.12376v2 Announce Type: replace-cross Abstract: When LLM conversations grow beyond the context window, old content must be evicted -- but ho…
5
EpiCache: Episodic KV Cache Management for Long-Term Conversation on Resource-Constrained Environments
EpiCache提出场景式KV缓存管理，在资源受限设备上高效支持超长对话，内存优化有新招。
arXiv:2509.17396v4 Announce Type: replace Abstract: Modern large language models (LLMs) extend context lengths to millions of tokens, enabling coheren…