文章预览
LG - 机器学习 CV - 计算机视觉 CL - 计算与语言 1、[LG] From Medprompt to o1：Exploration of Run-Time Strategies for Medical Challenge Problems and Beyond 2、[LG] LoRA vs Full Fine-tuning：An Illusion of Equivalence 3、[CL] Interpretable Language Modeling via Induction-head Ngram Models 4、[CL] Attacking Vision-Language Computer Agents via Pop-ups 5、[CL] Evaluation data contamination in LLMs：how do we measure it and (when) does it matter? 摘要：面向医疗挑战问题的运行时策略探索、LoRA与完全微调的显著不同、通过归纳头Ngram模型实现可解释语言建模、用弹窗攻击视觉语言计算机操纵Agent、大语言模型数据污染评估 1、[LG] From Medprompt to o1: Exploration of Run-Time Strategies for Medical Challenge Problems and Beyond H Nori, N Usuyama, N King, S M McKinney... [Microsoft & OpenAI] 从Medprompt到o1：面向医疗挑战问题的运行时策略探索 要点: o1-preview 的卓越性能：OpenAI 的 o1-prev
………………………………
