Brief

Researchers have developed a new auditing protocol for weak-label benchmarks in natural language processing. This protocol distinguishes between outputs predictable from metadata alone and those genuinely dependent on the provided evidence. By combining a metadata prior dominance score with an evidence intervention statistic, the method aims to provide a more robust evaluation of benchmark reliability. AI

Researchers have developed CANTANTE, a new framework designed to optimize the configuration of large language model-based multi-agent systems. This system addresses the challenge of assigning credit for performance when only system-level scores are available, by decomposing rewards into per-agent update signals. CANTANTE was evaluated on programming, mathematical reasoning, and question-answering tasks, where it demonstrated superior performance compared to existing methods and unoptimized prompts, while also incurring lower inference costs. AI