Announcement_13

Our paper TwinRouterBench, a benchmark for evaluating LLM routing in multi-step agent workflows, was accepted at NeurIPS 2026.