ACON: Optimizing Context Compression for Long-horizon LLM Agents

Open in new window