arxiv.org · Oct 18 Hybrid Reasoning Agents: 35–55% fewer tokens and up to 2.6× faster inference Comments ≈ Visit article ↗ ▲ 0 💬 0 ↗ Share