Skip to main content

NULL Pointer Dereference

The product dereferences a pointer that it expects to be valid but is NULL.

CWE-476 at MITRE →

24.1%

repair rate

14–33

95% interval

29

scored attempts · 2 codebases

By model

#AgentModelRepair rate with intervalRateAttempts
1codexgpt-5.2100.0% — no interval: too few codebases or attempts to generalise100.0%2
2claudeclaude-opus-4-650.0% — no interval: too few codebases or attempts to generalise50.0%2
3cursorgpt-5.3-codex50.0% — no interval: too few codebases or attempts to generalise50.0%2
4cursoropus-4.650.0% — no interval: too few codebases or attempts to generalise50.0%2
5gemini31gemini-3.1-pro-preview50.0% — no interval: too few codebases or attempts to generalise50.0%2
6opencodeclaude-opus-4-650.0% — no interval: too few codebases or attempts to generalise50.0%2
7claudeclaude-opus-4-50.0% — no interval: too few codebases or attempts to generalise0.0%2
8codexgpt-5.2-codex0.0% — no interval: too few codebases or attempts to generalise0.0%2
9cursorcomposer-1.50.0% — no interval: too few codebases or attempts to generalise0.0%2
10cursorgpt-5.20.0% — no interval: too few codebases or attempts to generalise0.0%2
11geminigemini-3-pro-preview0.0% — no interval: too few codebases or attempts to generalise0.0%2
12opencodeclaude-opus-4-50.0% — no interval: too few codebases or attempts to generalise0.0%1
13opencodegemini-3.1-pro-preview0.0% — no interval: too few codebases or attempts to generalise0.0%2
14opencodegpt-5.20.0% — no interval: too few codebases or attempts to generalise0.0%2
15opencodegpt-5.2-codex0.0% — no interval: too few codebases or attempts to generalise0.0%2

outcome mix

7 repaired · 15 not repaired · 7 did not build

7 repaired · 15 not repaired · 7 did not build

Corpus cve-bench-136 · run February 2026 · How we test

All weakness classes →