Route to Remember: Adaptive Memory Selection for LLM Agents
Sets the starting problem: text memory extracted from overflow context is a retrieval bottleneck.
3 items · 2 sources · 3 days
Operational story trace
Follow in this browser to see new updates on your Live feed.
Latest change
A Nov 19 paper reports deployment lessons from agents that distill memory from their own history to avoid repeat procedural failures such as mis-paginating an API or resolving the wrong entity.
This is a loose grouping of three independent arXiv papers on LLM agent memory, linked by topic rather than one event. It opened on Nov 3 with a paper on choosing which text memories to retrieve, then moved to memory for long-running character dialogue on Nov 12.
Arc
Sets the starting problem: text memory extracted from overflow context is a retrieval bottleneck.
Extends the theme from task agents to conversational memory that tracks evolving relationships.
Adds deployment evidence: memory learned from the agent's own history, aimed at named procedural failure modes.
Sets the starting problem: text memory extracted from overflow context is a retrieval bottleneck.
Extends the theme from task agents to conversational memory that tracks evolving relationships.
Adds deployment evidence: memory learned from the agent's own history, aimed at named procedural failure modes.
What to watch — open questions
Storylines are threaded mechanically from the feed: stories that share a distinctive anchor across multiple days and sources. Each item links to its original source. The evidence trace, current state, and open questions are written by the editor routine and refreshed whenever a new beat lands.