Tag
This paper from Stanford, NVIDIA, and UC Berkeley introduces LLM-as-a-Verifier, a training-free verification framework that uses continuous scoring from LLM logits to improve accuracy across coding, robotics, and medical domains, achieving state-of-the-art results on multiple benchmarks.
FrontierCode is a new benchmark for coding agents, human-verified with a continuous scoring model, designed to evaluate real-world performance.