Tag
论文提出「推导鸿沟」(derivation gap)现象:LLM 工具调用评审器会因输入中缺少可参考的原文,而拒绝接受正确但需推导或为默认值的参数。作者提出 DeRef 方法,通过独立模型生成参考答案来显著提升评审的 AUROC 表现。
TechCrunch revealed the five investor judges who will pick the winner of Startup Battlefield 200 at Disrupt 2026, running October 13-15. The panel includes Stacy Brown-Philpot of Cherryrock Capital and Chi-Hua Chien of Goodwater Capital, among others, with a live AMA to follow the final pitch.
TechCrunch has revealed the next wave of venture capitalists who will judge the Startup Battlefield 200 at Disrupt 2026, highlighting their role in evaluating early-stage startups at the conference.