Update README.md

README.md

@@ -8,23 +8,17 @@ pinned: false
 ---
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/kerneldf/kerneldf.github.io/main/og.png" alt="KernelDF — Data processing, down to the kernel." width="100%" />
+  <img src="https://raw.githubusercontent.com/kerneldf/kerneldf.github.io/main/huggingface/banner.png" alt="KernelDF" width="100%" />
 </p>
 
-`AI × data systems × GPU kernels`
-
-# Data processing, down to the kernel.
-
-KernelDF is a home for research and open-source software that uses AI to accelerate DataFrame operations and database query execution through specialized GPU kernels.
+Research and open-source software that uses AI to accelerate DataFrame operations and database query execution through specialized GPU kernels.
 
 **[Website](https://kerneldf.github.io)** · **[GitHub](https://github.com/kerneldf)** · **[Datasets](https://huggingface.co/kerneldf/datasets)**
 
 ---
 
-## DataKernelBench
-
-*Can LLMs optimize database queries on GPUs?*
+### DataKernelBench
 
-A benchmark for translating SQL-derived PyTorch programs into optimized CUDA and Triton kernels, evaluated for correctness and execution speed. Accepted at EMNLP 2026.
+A benchmark for translating SQL-derived PyTorch programs into optimized CUDA and Triton kernels. Accepted at EMNLP 2026.
 
-[Dataset](https://huggingface.co/datasets/kerneldf/datakernelbench) · [Paper](https://arxiv.org/abs/2608.25061) · [Code](https://github.com/kerneldf/DataKernelBench) · [Project page](https://kerneldf.github.io/datakernelbench/)
+[Dataset](https://huggingface.co/datasets/kerneldf/datakernelbench) · [Paper](https://arxiv.org/abs/2608.25061) · [Code](https://github.com/kerneldf/DataKernelBench)