DEFault++: Automated Fault Detection, Categorization, and Diagnosis for Transformer Architectures
Sigma Jahan [email protected], Saurabh Singh Rajput [email protected], Tushar Sharma [email protected], Mohammad Masudur Rahman [email protected] · 2026
Read DEFault++: Automated Fault Detection, Categorization, and Diagnosis for Transformer Architectures by Sigma Jahan [email protected], Saurabh Singh Rajput [email protected], Tushar Sharma [email protected], Mohammad Masudur Rahman [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-28118.txt.
Introduction
Jahan's team presents DEFault++, a hierarchical technique for finding and tracing faults in transformer models that quietly change conduct without runtime errors. It first detects a fault, then names the faulty component, then the cause within it, using a fault spread graph over component readings. Trained on a benchmark of thousands of mutant-testing runs, it beats four prior methods on detection. It also lifts developer repair accuracy sharply in a user study.
Catalog record
| CREATOR | Sigma Jahan [email protected], Saurabh Singh Rajput [email protected], Tushar Sharma [email protected], Mohammad Masudur Rahman [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 169698 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.28118