In-Model Context Compression with Agentic Soft Tokens for Long-Horizon LLM Agents
This paper introduces Agentic Soft-Token Compression (ASTC), a training-free compression method using soft tokens within LLM agent for long-horizon tasks that matches or exceeds the uncompressed agent's accuracy with up to 50.68% fewer end-to-end Attention Pairs.
Xu-Ning He, Kai-Hsiang Fan, Run-Yu Zhang et al.
· 0 citations