Open Source Automated Interpretability for Sparse Autoencoder Features
AI 摘要：Building and evaluating an open-source pipeline for au...
The Practitioner’s Guide to the Maximal Update Parameterization
本文源自EleutherAI Blog的研究专栏，聚焦最大更新参数化（Maximal Update Parameteri...
Mathematical proof reveals why fixed AI guardrails can never block every jailbreak
AI 摘要：Can we make artificial intelligence impervious to adve...
Osprey-inspired algorithm lifts Chinese-English translation accuracy
AI 摘要：Research published in the International Journal of Inf...
Robots are closing in on human-like judgments, addressing a key challenge in physical AI
AI 摘要：KAIST researchers solved a key challenge in the commer...
‘Reading relationships, crunching stats’—184-times faster data analysis
AI 摘要：A research team at POSTECH led by Professor Wook-Shin ...
How AI chatbots become better learning coaches
AI 摘要：Many AI systems answer questions in a matter of second...
Thirty-five AI comedians walked into a workshop, and what happened next could reshape how machines learn humor
来自Tech Xplore AI的相关研究显示，乔治梅森大学博士生Shiwei Hong受人类编剧集体工坊的启发，发起了...
Researchers develop AI-powered railway control system for efficient urban train operation
AI 摘要：As cities continue to expand, railways are expected to...