Update arXiv citation

CITATION.cff

@@ -26,3 +26,25 @@ authors:
 message: "If you use this software, please cite it using these metadata."
 date-released: 2020-09-03
 repository-code: "https://github.com/pytorch/benchmark"
+preferred-citation:
+  type: article
+  authors:
+  - family-names: "Hao"
+    given-names: "Yueming"
+  - family-names: "Zhao"
+    given-names: "Xu"
+  - family-names: "Bao"
+    given-names: "Bin"
+  - family-names: "Berard"
+    given-names: "David"
+  - family-names: "Constable"
+    given-names: "Will"
+  - family-names: "Aziz"
+    given-names: "Adnan"
+  - family-names: "Liu"
+    given-names: "Xu"
+  title: "TorchBench: Benchmarking PyTorch with High API Surface Coverage"
+  year: 2023
+  eprint: "2304.14226"
+  primary-class: "cs.LG"
+  archivePrefix: "arXiv"  

README.md

@@ -145,3 +145,18 @@ See [Unidash](https://www.internalfb.com/intern/unidash/dashboard/pytorch_benchm
 ## Adding new models
 
 See [Adding Models](torchbenchmark/models/ADDING_MODELS.md).
+
+## Citation
+
+If you use TorchBench in your academic work, please cite the following:
+```
+@misc{hao2023torchbench,
+      title={TorchBench: Benchmarking PyTorch with High API Surface Coverage}, 
+      author={Yueming Hao and Xu Zhao and Bin Bao and David Berard and Will Constable and Adnan Aziz and Xu Liu},
+      year={2023},
+      eprint={2304.14226},
+      archivePrefix={arXiv},
+      primaryClass={cs.LG}
+}
+```
+The full paper [TorchBench: Benchmarking PyTorch with High API Surface Coverage](https://arxiv.org/abs/2304.14226) is public on arXiv.org.