---
format: "aidr-story-markdown/v1"
id: "eaf0a1df2e81af460cdcd09f9192661139315b2d9e43826a974b98750438b907"
canonical_url: "https://aidr.today/eaf0a1df?lang=en"
title: "A Team of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-22T01:51:43.000Z"
category: "Research"
topics: ["safety"]
source_urls: ["https://www.marktechpost.com/2026/11/21/columbia-certik-tensorlean-formal-verification-gpu-kernels/"]
summary: "Researchers from Columbia University and CertiK introduced TensorLean, which formally verifies production GPU kernels at the PTX level using Lean. The tool checked 149 kernel instances, proving 83 correct and finding 38 bugs in FlashAttention, DeepGEMM, and others."
---

# A Team of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels

> [Open the canonical story](<https://aidr.today/eaf0a1df?lang=en>)

**Published:** 2026-11-22T01:51:43.000Z
**Category:** Research
**Topics:** safety

## Summary

Researchers from Columbia University and CertiK introduced TensorLean, which formally verifies production GPU kernels at the PTX level using Lean\. The tool checked 149 kernel instances, proving 83 correct and finding 38 bugs in FlashAttention, DeepGEMM, and others\.

## Sources

- [Story source](<https://www.marktechpost.com/2026/11/21/columbia-certik-tensorlean-formal-verification-gpu-kernels/>)

