gemini-4IMAGE
dfbench 实测 Gemini 4 Argon：漏洞召回率 74%
depthfirstlabs 在 dfbench 漏洞检测基准上测 Gemini 4 Argon，应用与智能合约任务召回 74%，165 项总分 67.5%，仅次于 Mythos 5。
Ram
@PothamRam15 天前收录原推发布于 2026.10.20
Ram @PothamRam原推发布于 2026.10.20
We tested Gemini 4 Argon on @depthfirstlabs dfbench, our benchmark for finding real vulnerabilities in code. On application and smart-contract tasks, where it rarely refused, it found 74% of known vulnerabilities, the highest recall we have measured. Across all 165 tasks it scored 67.5%, second only to Mythos 5 (69%{.
x.com/pothamram/status/2112309211854180666采集于 2026.10.20 09:59