Automated AI Research Agent Finds Numerical Bug in FlashInfer Kernels
Josh Tobin's team reported that their automated research agent, using a previously built reward-hacking judge, uncovered a numerical bug in FlashInfer kernels used by vLLM and SGLang—hardcoded -50000 attention mask values—and its fix was merged by experts.
2026-08-28 ~ 2026-08-28 · 3 related posts
- AI automated research finds numerical bug in vLLM/SGLang backend — josh_tobin_ · 2026-08-28
- Automated research agent finds silent numerical bug in FlashInfer kernels; fix PR accepted — timshi_ai · 2026-08-28
- Auto-Research System Finds Numerical Bug in vLLM/SGLang Kernels — PMinervini · 2026-08-28