New research from Meta and CMU.
This one is on agentic context management for long horizon tasks.
（bookmark it）
Production agents accumulate context every turn. The usual fix compresses on a token threshold and throws the remainder away， so the trigger sometimes fires for reasons unrelated to what the agent is currently working on.
ACM hands the agent purpose-built context editing tools. It decides when to compress， offloads what it drops into an external memory store， and queries that store later when it needs the detail back. Compression shifts from short-term to long-term memory.
They also build a post-training pipeline on high-quality context management demonstrations. That yields a 27% relative gain on BrowseComp-Plus and gets close to open-source models 40 times larger.
Better context management lowers peak token pressure， lets the agent explore longer before running out of room， and produces more consistent solutions across independent trials of the same task.
Code， data， and checkpoints are released.
Paper： https://arxiv.org/abs/2607.23809
Learn to build effective AI agents in our academy： https://academy.dair.ai/