Focus on What Matters: Fisher-Guided Adaptive Multimodal Fusion for Vulnerability Detection CCS: Software and its engineering Software testing and debugging CCS: Security and privacy Software security engineering CCS: Computing methodologies Neural networks CCS: Software and its engineering Software
Yun Bian, Yi Chen, HaiQuan Wang, Shihao Li, Zhe Cui · 2026
Read Focus on What Matters: Fisher-Guided Adaptive Multimodal Fusion for Vulnerability Detection CCS: Software and its engineering Software testing and debugging CCS: Security and privacy Software security engineering CCS: Computing methodologies Neural networks CCS: Software and its engineering Software by Yun Bian, Yi Chen, HaiQuan Wang, Shihao Li, Zhe Cui (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-2601-02438.txt.
Description
Yun Bian and four co authors question a common recipe for detecting security flaws in code, which fuses two views of a program. Its token sequence as read by a pre trained model and its code property graph as read by a graph network. The recipe assumes a second view always adds information. Their analysis shows pre trained models already encode much structural information, so the two views overlap heavily, and graph encoders are usually weaker feature extractors.
Naive fusion can therefore dilute the strongest warning cues with noise instead of sharpening them. Their answer is a task conditioned complementary fusion strategy guided by Fisher information, a measure of how much a feature tells about the label. Fusion then keeps what the second view genuinely adds for the vulnerability task, rather than blending everything by default.
Catalog record
| CREATOR | Yun Bian, Yi Chen, HaiQuan Wang, Shihao Li, Zhe Cui |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-02438