Reva: Real-Time Vulnerability Detection through Contextual Probing of Large Language Models
arXiv cs.AI · 27 Oct 2026, 03:50 AEST
REVA is an AI vulnerability detector that enhances real-time screening of code functions. It achieves high accuracy while significantly reducing processing time compared to existing methods.
Key facts
- A vulnerability costs more to fix the later it is discovered.
- REVA (REal-time Vulnerability Analysis) is an AI vulnerability detector that brings the security context and code understanding components behind agentic detectors into a single forward pass of an LLM.
- On PrimeVul's held-out test set, REVA on GLM-5.3-Flash reaches 60.0% accuracy in about one second per function.
- REVA is over 250 times faster and less than a tenth as costly as the state-of-the-art agentic detector, VulTriage.
- Across three open LLMs, REVA is at least as accurate as VulTriage, and about 10% more accurate on two of them.
Read aloud
REVA is an A I vulnerability detector that enhances real-time screening of code functions. It achieves high accuracy while significantly reducing processing time compared to existing methods. A vulnerability costs more to fix the later it is discovered. REVA REal-time Vulnerability Analysis is an A I vulnerability detector that brings the security context and code understanding components behind agentic detectors into a single forward pass of an L L M. On PrimeVul's held-out test set, REVA on GLM-five point three-Flash reaches sixty point zero percent accuracy in about one second per function. REVA is over two hundred fifty times faster and less than a tenth as costly as the state-of-the-art agentic detector, VulTriage. Across three open L L M s, REVA is at least as accurate as VulTriage, and about ten percent more accurate on two of them.
Sources in this cluster
- Reva: Real-Time Vulnerability Detection through Contextual Probing of Large Language Models
arXiv cs.AI · 27 Oct 2026, 03:50 AEST