Model2Kernel: Model-Aware Symbolic Execution For Safe CUDA Kernels

arXiv 2603.24595•dcbb2cbb7eab95d3a01e88b96dabf58e0ae30065ca116e1922b3822e3c85d318
AI-safetyCUDAGPULLM-inferenceRAGdynamic-analysisformal-verificationmemory-corruptionmemory-safetymodel-securityprogram-synthesisself-evolving-agentsstatic-analysissymbolic-executiontrusted-AIvulnerability-discovery

Paper metadata

arXiv ID
2603.24595
Version
Not specified by this published record
Category
Computer Science — Programming Languages (cs.PL)

The PDF link points to arxiv.org. Baitaphish does not expose a private stored PDF.

Evidence and limitations

Source ID
arxiv_cs_pl
Record identifier
dcbb2cbb7eab95d3a01e88b96dabf58e0ae30065ca116e1922b3822e3c85d318
Enrichment time
2026-03-27T08:52:01Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.

Model2Kernel: Model-Aware Symbolic Execution For Safe CUDA Kernels · Baitaphish