Auto-Research System Finds Numerical Bug in vLLM/SGLang Kernels
PMinervini · x · 2026-08-28
An auto-research system at RecursiveSI discovered that some FlashInfer kernels (underpinning vLLM and SGLang) used a hardcoded value of -50,000 as a masked-attention sentinel, which can be smaller than valid QK values and cause instability. The edge case was identified and has been fixed. Silent numerical bugs like this can cause significant debugging headaches.
Related event: Automated AI Research Agent Finds Numerical Bug in FlashInfer Kernels(3 posts)→
More from Infra
- Samsung cuts Q4 phone output as Nittobo shifts E-glass to AI-chip T-glass — zephyr_z9 · 2026-10-16
- White House AI chief discusses Empire AI, New York's shared compute infrastructure — mkratsios47 · 2026-10-16
- FT warns of 'Darwin moment' for AI borrowers as CoreWeave bonds sink to distress levels — firstadopter · 2026-10-16
- Phone as interface, home servers as the lab: a two-workflow local AI setup — EAccelerate_42 · 2026-10-16
- "A datacenter is a bank": CoreWeave drama revives calls for算力 supervision — arthurcolle · 2026-10-16
- DOT-MoE: converting dense models into sparse MoE via optimal transport — burkov · 2026-10-16