Understand this story
On November 3, 2026, Cogent Security Inc. released vulnerability remediation model VR-1.1, used to write remediation patches for newly disclosed software vulnerabilities. According to Unite.AI, among the 190 tasks in LiveCVEBench, VR-1.1 achieved a pass remediation rate of 57.9%, while Claude Code and Codex each achieved 45.3%. The benchmark covers 14 programming languages and 153 repositories, and requires patches to pass both vulnerability checks and functionality checks. SiliconANGLE reported that the benchmark was built by Harbin Institute of Technology and Kuaishou, and that Cogent cut off external network access during remediation and hid the scoring tests until each agent had finished. Cogent stated that the results come from security remediation training, as well as a process of first reproducing the vulnerability, then applying the patch and detecting bypasses; it also noted that the baseline models' training cutoff dates are later than the publication dates of the relevant CVEs, and that production deployment still requires review and testing in the target environment.
AI-generated from public reports
Latest development · Report
Cogent releases VR-1.1, achieving a 57.9% fix pass rate on LiveCVEBenchDevelopments
1 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 3
Unite.AI
Cogent releases VR-1.1, achieving a 57.9% fix pass rate on LiveCVEBench
Cogent Security released the vulnerability-fixing model VR-1.1 on November 3, 2026, achieving a 57.9% pass fix rate across 190 tasks in LiveCVEBench, compared with 45.3% each for Claude Code and Codex. The benchmark covers 14 programming languages and 153 repositories, and requires patches to pass both vulnerability checks and functionality checks. Cogent says the results come from security-fix training and a process of first reproducing the vulnerability, then applying a patch and detecting bypasses, while also noting that the baseline models' training cutoff dates are later than the publication dates of the relevant CVEs, and that production deployment still requires review and testing in the target environment.
SiliconANGLE
Cogent Security releases VR-1.1 vulnerability-fixing model, scoring 57.9% on LiveCVEBench, surpassing…
Cogent Security has released VR-1.1, a frontier reasoning model designed to write fix patches for newly disclosed software vulnerabilities. On the 190 tasks in LiveCVEBench, VR-1.1 produced usable fixes for 57.9% of tasks, while Claude Code and Codex both scored 45.3%. The benchmark was built by Harbin Institute of Technology and Kuaishou, covers 14 programming languages, and requires patches to pass both security tests and functional tests; Cogent cut off external network access during fixing and hid the scoring tests until each agent had finished.
Heat trend for this story
- Current
- 11
- Period peak
- 18 Nov 4 02:00
- 24-hour change
- —
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.