【直播回放】NeurIPS 2025精选论文解读

直播回放：
-Controlling Thinking Speed in Reasoning Models
林政楷 浙江大学

-Data-Juicer 2.0: Cloud-Scale Adaptive Data Processing for and with Foundation Models
陈道源 通义实验室研究员

-Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decoding
王一鸣 上海交通大学

-UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface
汤 昊 北京大学

-Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
邱子涵 通义实验室算法科学家