Cache Engineering
An optimization approach for long-trajectory reasoning in LLM agents that directly manipulates the active KV cache (such as discarding or relocating cache entries) to avoid re-prefilling, thereby reducing the cost and latency of long-context tasks.
Timeline (last 90 days)
A paper proposes a new approach called "Cache Engineering": letting the agent directly manipulate the active KV cache without modifying the trajectory or re-prefilling.
Cache engineering manages the KV cache like memory and is suited to tasks that require multiple rounds of tool calls and long-horizon planning.
All Facts (2)
A paper proposes a new approach called "Cache Engineering": letting the agent directly manipulate the active KV cache without modifying the trajectory or re-prefilling.
50%UnverifiedCache engineering manages the KV cache like memory and is suited to tasks that require multiple rounds of tool calls and long-horizon planning.
50%