Newest papers

RouterWorld: Scaling Conversational AI with Latency-Aware LLM Routing

Jinyang Li, Denis Emelin, Simone Tedeschi, Weiwei Cheng, Cristina Morariu, Haytham Assem, Jens Lehmann, Emine Yılmaz, Angus Addlesee · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26511 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Sin embargo, los estándares de enrutamiento de LLM existentes asumen la plena observabilidad de todos los resultados del modelo de candidato, subestiman la latencia como una métrica de primera clase, y la evaluación de fragmentos a través de protocolos heterogéneos desalineados con los ajustes del mundo real. Introducimos RouterWorld, un benchmark integral para el enrutamiento de LLM en el GenAI de conversación. RouterWorld unifica las respuestas de un solo giro de 371k y las trayectorias de agentes de varios giros de 76k, adopta las interacciones naturales del usuario con un protocolo de juicio LLM estándar, y descubre la eficiencia del precio de transición del público a través de anuncios de latencia basados en hardware. Crucialmente, RouterWorld formaliza el enrut

Machine translation (Meta M2M100); the English original follows.

Abstract

Conversational AI systems built on Large Language Models (LLMs) require routing mechanisms that balance answer quality and response latency at scale. However, existing LLM routing benchmarks assume full observability of all candidate model outcomes, under-emphasize latency as a first-class metric, and fragment evaluation across heterogeneous protocols misaligned with real-world settings. We introduce RouterWorld, a comprehensive benchmark for LLM routing in conversational GenAI. RouterWorld unifies 371k single-turn responses and 76k multi-turn agentic trajectories, adopts natural user interactions with a standardized LLM-as-a-judge protocol, and decouples efficiency from transient pricing via hardware-grounded latency annotations. Crucially, RouterWorld formalizes routing under partial observability through two setups: Cold-Start, which restricts learning to sparse, static logs, and Voyager, which models continuous data streams where routing decisions shape future training distributions. Evaluating 13 routing baselines, we find that most routers over-emphasize accuracy while ignoring latency under partial observability. Code and benchmark data are publicly available.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Observability, Latency (audio), Static routing, Benchmark (surveying), Link-state routing protocol

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange