Skip to content

Accurate, fast FP4 KV-cache decoding through algorithm-system co-design.

docs: add arXiv, license and OpenBitSys badges and citation

DD-DuDa4e2907d74d8b
diff --git a/README.md b/README.md
index 495612e..32e4bfe 100644
--- a/README.md
+++ b/README.md
@@ -2,6 +2,12 @@
<img src="imgs/title.svg" alt="BitDecoding-FP4" width="550">
</p>
+<p align="center">
+ <a href="https://arxiv.org/abs/2611.09825"><img src="https://img.shields.io/badge/arXiv-2611.09825-b31b1b.svg" alt="arXiv:2611.09825"></a>
+ <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-green.svg" alt="License: MIT"></a>
+ <a href="https://github.com/OpenBitSys"><img src="https://img.shields.io/badge/OpenBitSys-181717.svg?logo=github&amp;logoColor=white" alt="OpenBitSys"></a>
+</p>
+
BitDecoding-FP4 delivers accurate, high-performance LLM decoding through
**algorithm-system co-design**, combining **FP4 KV-cache quantization** with
native FP4 attention kernels. It integrates into vLLM as an out-of-tree `CUSTOM`
@@ -62,3 +68,19 @@ See the [accuracy evaluation guide](evaluation/README.md) for dataset installati
Original code is licensed under the [MIT License](LICENSE).
Third-party components retain their own licenses; see [NOTICE](NOTICE).
+
+## Citation
+
+If you use BitDecoding-FP4 in your research, please cite:
+
+```bibtex
+@misc{du2026bitdecodingfp4nativefp4kv,
+ title={BitDecoding-FP4: Native FP4 KV Cache Decoding via Algorithm-System Co-Design},
+ author={Dayou Du and Woon Yee Ng and Guangming Sheng and Nick Lau and Jianyi Cheng and Luo Mai},
+ year={2026},
+ eprint={2611.09825},
+ archivePrefix={arXiv},
+ primaryClass={cs.AI},
+ url={https://arxiv.org/abs/2611.09825},
+}
+```