Chain-of-thought monitorability could improve generative AI safety by assessing how models come to their conclusions and ...

Deep research’ AI agents combine large language models with sophisticated reasoning frameworks to conduct in-depth, ...

Assessing the progress of new AI language models can be as challenging as training them. Stanford researchers offer a new approach.

When computer-based tools are used for analysis of complex systems, the design of user interactions and interfaces becomes an essential part of development that determines the overall quality. The ...

To illustrate how to select the optimal process configuration using a sustainability analysis that incorporates the F&EI as a safety indicator, this paper presents a representative case study. The ...

This paper presents a process-oriented GCC failure analysis method based on fault injection. We first analyze the process of GCC, systematically define the fault modes of GCC, and establish a GCC ...