[State of Evals] LMArena's $1.7B Vision — Anastasios Angelopoulos, LMArena

Arena’s central measurement is whether people prefer an AI response to a question they actually wanted answered. In the interview, the model-comparison platform’s cofounder argues that letting users supply their own requests gives its votes a connection to everyday use that preset questions can miss.

That makes the audience part of the measurement. The company describes using surveys and prompt analysis to understand who participates, while acknowledging survey-response bias. A ranking therefore reflects preferences among the people and tasks reaching the platform; understanding that population helps explain what the score represents.