一作ICLR 2026Oral· Deep Learning / AlgorithmsReasoning with Sampling: Your Base Model is Smarter Than You ThinkAayush Karan · Yilun Du推理OpenReview