LATTICE: Evaluating Decision Support Utility of Crypto Agents
Abstract:We introduce LATTICE, a benchmark for evaluating the decision support utility of crypto agents in realistic user-facing scenarios. Crucially, the dimensions and tasks are designed to be evaluable at scale using LLM judges, without relying on ground truth from expert annotators or external data sources. LATTICE addresses this gap by: (1) defining six evaluation dimensions that capture key decision support […]