An 8B model trained with SFT and GRPO reinforcement learning generates Triton kernels that match or exceed frontier model performance on TritonBench and KernelBench correctness metrics.
Program synthesis using deduction-guided reinforcement learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AutoTriton: Automatic Triton Programming with Reinforcement Learning in LLMs
An 8B model trained with SFT and GRPO reinforcement learning generates Triton kernels that match or exceed frontier model performance on TritonBench and KernelBench correctness metrics.