PACE method targets KV cache limits for long-horizon LLM agents
A new arXiv paper, PACE, argues that keeping the right tokens in the KV cache is insufficient for long-horizon LLM agents, whose sessions can reach millions of tokens and exceed GPU memory.
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-11-10 05:00 · arXiv cs.LG
PACE: Keeping the Right Tokens in the KV Cache Is Not Enough for Long-Horizon LLM Agents