mlsanityData to 14 Oct 2026

OptiML: An End-to-End Framework for Program Synthesis and CUDA Kernel Optimization

arXiv 2602.12305 ↗First posted 12 Feb 2026cs.LG cs.AI cs.DCcited by 3

Not on the map. Why?

Abstract (arXiv)

Generating high-performance CUDA kernels remains challenging due to the need to navigate a combinatorial space of low-level transformations under noisy and expensive hardware feedback. Although large language models can synthesize functionally correct CUDA code, achieving competitive performance requires systematic exploration and verification of optimization choices. We present OptiML, an end-to-end framework that maps either natural-language intent or input CUDA code to performance-optimized CUDA kernels by formulating kernel optimization as search under verification. OptiML consists of two decoupled stages. When the input is natural language, a Mixture-of-Thoughts generator (OptiML-G) acts as a proposal policy over kernel implementation strategies, producing an initial executable program. A search-based optimizer (OptiML-X) then refines either synthesized or user-provided kernels using Monte Carlo Tree Search over LLM-driven edits, guided by a hardware-aware reward derived from profiler feedback. Each candidate transformation is compiled, verified, and profiled with Nsight Compute, and evaluated by a composite objective that combines runtime with hardware bottleneck proxies and guardrails against regressions. We evaluate OptiML in both synthesis-and-optimize and optimization-only settings on a diverse suite of CUDA kernels. Results show that OptiML consistently discovers verified performance improvements over strong LLM baselines and produces interpretable optimization trajectories grounded in profiler evidence.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 22 of its 35 references that are in the map. First posted on arXiv 12 Feb 2026.

Builds on

22 of 35 references.

References · 22

  1. 2017Attention Is All You Needcited by 60,361
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  3. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  4. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,673
  5. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,181
  6. 2024A Survey on Large Language Models for Code Generationcited by 658
  7. 2025KernelBench: Can LLMs Write Efficient GPU Kernels?cited by 225
  8. 2020Ansor: Generating High-Performance Tensor Programs for Deep Learningcited by 201
  9. 2025Monte Carlo Tree Search for Comprehensive Exploration in LLM-Based Automatic Heuristic Designcited by 109
  10. 2025Kevin: Multi-Turn RL for Generating CUDA Kernelscited by 78
  11. 2025TritonBench: Benchmarking Large Language Model Capabilities for Generating Triton Operatorscited by 75
  12. 2025AutoTriton: Automatic Triton Programming with Reinforcement Learning in LLMscited by 40
10 more
  1. 2025CUDA-LLM: LLMs Can Write Efficient CUDA Kernelscited by 34
  2. 2025MCTS-Judge: Test-Time Scaling in LLM-as-a-Judge for Code Correctness Evaluationcited by 32
  3. 2025Geak: Introducing Triton Kernel AI Agent & Evaluation Benchmarkscited by 32
  4. 2024CodeRosetta: Pushing the Boundaries of Unsupervised Code Translation for Parallel Programmingcited by 19
  5. 2025CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learningcited by 13
  6. 2025Mixture of Thoughts: Learning to Aggregate What Experts Think, Not Just What They Saycited by 12
  7. 2024HPC-Coder-V2: Studying Code LLMs Across Low-Resource Parallel Languagescited by 8
  8. 2025Prompt Alchemy: Automatic Prompt Refinement for Enhancing Code Generationcited by 8
  9. 2021LS-CAT: A Large-Scale CUDA AutoTuning Datasetcited by 3
  10. 2021Searching CUDA code autotuning spaces with hardware performance counters: data from benchmarks running on various GPU architecturescited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Optimizing CUDA like a Human: Micro-Profiling Tools as Expert Surrogates for LLM-Based GPU Kernel Optimizationcited by 4
  2. 2026Multi-turn RL with Structural and Performance Aware Rewards for CUDA Kernel Generationcited by 0
  3. 2026KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimizationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026CUDA-Harness: Harnessing Agentic CUDA Kernel Generation and Optimization from Natural Language
  2. 2025Towards Robust Agentic CUDA Kernel Benchmarking, Verification, and Optimization
  3. 2025From Large to Small: Transferring CUDA Optimization Expertise via Reasoning Graph
  4. 2025cuPilot: A Strategy-Coordinated Multi-agent Framework for CUDA Kernel Evolution
  5. 2026From Tokens to Regions: CUDA-Sensitive Instruction Tuning for GPU Kernel Generation
  6. 2025Evolution of Kernels: Automated RISC-V Kernel Optimization with Large Language Models
  7. 2025CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning
  8. 2026Kernel Forge: An Agent Harness for LLM-based Generation and Optimization of CUDA Kernels