.png){kind=icon}
.png){kind=icon}
Return
Automated robustness testing for LLM-based natural language processing software
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.130642.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Propose AORTA, an automated robustness testing framework for LLM-based NLP software. • Develop ABS, an efficient testing method using adaptive beam search and backtracking. • Achieve an average test validity of 86.1. • Reduce test cost by up to 3442s and 219 queries per case over the strongest baseline.
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W