Lilian Weng
PulseAugur coverage of Lilian Weng — every cluster mentioning Lilian Weng across labs, papers, and developer communities, ranked by signal.
1 天有情绪数据
-
前 OpenAI 研究员的 AI 愿景与中国公司此前的发布相呼应
前 OpenAI 研究员 Lilian Weng 的新公司 Thinking Machines Lab (TML) 揭示了全双工、实时对话式 AI 的愿景。这一概念与中国公司“面壁智能”三个月前开源的 MiniCPM-o 4.5 所展示的功能非常相似。TML 和“面壁智能”都旨在摆脱传统的轮流对话式 AI 交互，提出“全双工”或“时间对齐微回合”框架，以处理交错的多模态信息流。
-
OpenAI Board Establishes Safety and Security Committee Amidst AGI Push
OpenAI has established a new Safety and Security Committee composed of board members and key technical experts. This committee will evaluate and recommend critical safety and security decisions for all OpenAI projects a…
-
用于视频生成的扩散模型
研究人员正在探索用于视频生成的先进扩散模型，以解决时间一致性和数据稀缺性等挑战。新方法侧重于改进参数化，例如 v-prediction 技术，并结合条件采样来完成扩展视频长度或填充缺失帧等任务。同时，通过训练后框架、混合注意力机制和语义视觉适应性，也在努力提高效率和可控性，目标是实现实时生成和更高质量的输出。
-
Contrastive learning advances model robustness and transparency in AI
Contrastive learning is a machine learning technique that creates an embedding space where similar data points are grouped together and dissimilar ones are separated. This method can be applied in both supervised and un…
-
OpenAI分享模型部署在AI安全和滥用方面的经验教训
OpenAI分享了部署其语言模型的经验，强调实际滥用情况常与最初的担忧不同。该公司强调了当前评估方法的局限性，以及解决安全问题需要新的基准。OpenAI还指出，基础安全研究显著提高了AI系统的商业效用。
-
Optimizing Transformer Inference: Techniques for Faster, Cheaper Large Models
Large transformer models present significant inference challenges due to their substantial memory footprint and computation costs, which scale quadratically with input length. Researchers and practitioners are exploring…
-
Researchers advance Bayesian Optimization for efficient decision-making and hyperparameter tuning
Several recent arXiv papers explore advancements in multi-armed bandit problems, a framework for sequential decision-making under uncertainty. Research includes handling changing action availability with "Flickering Mul…