# M2K: Making the Model-Kernel Interface Explicit for Reliable CUDA Kernel Verification

arXiv: https://arxiv.org/abs/2603.24595 · HTML: https://mlsanity.com/paper/2603.24595/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 78 references that are in the map.
- First posted on arXiv 6 Mar 2026.

## Builds on (most cited first, 26 of 26)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [Sparks of Artificial General Intelligence: Early experiments with GPT-4](https://mlsanity.com/paper/2303.12712/) (2023)
- [Solving Quantitative Reasoning Problems with Language Models](https://mlsanity.com/paper/2206.14858/) (2022)
- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)
- [A Survey on Large Language Models for Code Generation](https://mlsanity.com/paper/2406.00515/) (2024)
- [A Survey on Large Language Models for Recommendation](https://mlsanity.com/paper/2305.19860/) (2023)
- [Large Language Models for Information Retrieval: A Survey](https://mlsanity.com/paper/2308.07107/) (2023)
- [A Complete Survey on LLM-based AI Chatbots](https://mlsanity.com/paper/2406.16937/) (2024)
- [A Survey of Large Language Models for Code: Evolution, Benchmarking, and Future Trends](https://mlsanity.com/paper/2311.10372/) (2023)
- [PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression](https://mlsanity.com/paper/2405.14852/) (2024)
- [Any-Precision LLM: Low-Cost Deployment of Multiple, Different-Sized LLMs](https://mlsanity.com/paper/2402.10517/) (2024)
- [NNSmith: Generating Diverse and Valid Test Cases for Deep Learning Compilers](https://mlsanity.com/paper/2207.13066/) (2022)
- [ShiftAddLLM: Accelerating Pretrained LLMs via Post-Training Multiplication-Less Reparameterization](https://mlsanity.com/paper/2406.05981/) (2024)
- [PyCG: Practical Call Graph Generation in Python](https://mlsanity.com/paper/2103.00587/) (2021)
- [ProofWright: Towards Agentic Formal Verification of CUDA](https://mlsanity.com/paper/2511.12294/) (2025)
- [CUDA Leaks: Information Leakage in GPU Architectures](https://mlsanity.com/paper/1305.7383/) (2013)
- [Buffer overflow vulnerabilities in CUDA: a preliminary analysis](https://mlsanity.com/paper/1506.08546/) (2015)
- [MirrorFuzz: Leveraging LLM and Shared Bugs for Deep Learning Framework APIs Fuzzing](https://mlsanity.com/paper/2510.15690/) (2025)
- [GPUArmor: A Hardware-Software Co-design for Efficient and Scalable Memory Safety on GPUs](https://mlsanity.com/paper/2502.17780/) (2025)
- [CuFuzz: Hardening CUDA Programs through Transformation and Fuzzing](https://mlsanity.com/paper/2601.01048/) (2026)
- [Challenges and Design Considerations for Finding CUDA Bugs Through GPU-Native Fuzzing](https://mlsanity.com/paper/2603.05725/) (2026)

## Abstract (arXiv)

Large language model (LLM) inference systems rely on CUDA kernels for core GPU computations, yet the interface between models and kernels is implicit and poorly specified. Models and kernels evolve independently and often make incompatible assumptions about tensor shapes and input sizes, leading to subtle memory bugs in CUDA kernels. These bugs can crash inference services, corrupt model weights, or be exploited by remote adversaries. Existing techniques either incur prohibitive runtime overhead, require specialized hardware, or fail to handle dynamic tensor shapes and variable kernel launch configurations, leaving the CUDA memory bugs largely unaddressed. This paper presents M2K, a fully automated framework that makes the model-kernel interface explicit and leverages it to detect memory bugs in CUDA kernels used in LLM inference systems. M2K consists of two components. HFProbe traces model execution without GPU hardware, classifies kernel arguments into model-fixed and user-variable, and emits symbolic constraints that capture the interface. cuKLEE then performs symbolic execution on CUDA kernels to pinpoint memory bugs under the interface constraints, modeling tensors as disjoint memory regions and treating thread identifiers symbolically to scale to thousands of threads. In the evaluation, M2K discovers 181 previously unknown bugs in real LLM inference systems, while producing only nine false positives, demonstrating its effectiveness.