ResearchTasksMulti-Head Self-AttentionMulti-Head Self-AttentionFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedCUDA-LLM kernels task suite Multi-Head Self-AttentionOptiML6.85Execution Time (s)5Feb 27, 2026CUDA-LLM kernels task suite 1.0 (test)OptiML6.85Execution Time5Feb 27, 2026CUDA-LLM kernels task suite (test)——Latency (s)0Feb 26, 2026