AI RESEARCH
Adaptive Latent Agentic Reasoning
arXiv CS.CL
•
ArXi:2606.02871v1 Announce Type: new Large reasoning models improve performance by generating extended chain-of-thought (CoT) reasoning, but this behavior becomes inefficient when applied to LLM agents. Current LLM agents often generate verbose textual reasoning at every decision step and allocate reasoning effort nearly uniformly across turns, leading to substantial inefficiency in multi-turn agentic trajectories.