Update README.md

README.md

@@ -27,4 +27,38 @@ configs:
     - "real.parquet"
 ---
 
-The proof of the seed problems and the evaluation scripts can be found at https://github.com/haoyuzhao123/LeanIneqComp
+<div align="center">
+    <h1> <a href="https://arxiv.org">Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving on Inequalities</a></h1>
+</div>
+
+<div align="center">
+
+[![GitHub](https://img.shields.io/badge/GitHub-181717?style=flat&logo=github&logoColor=white)](https://github.com/haoyuzhao123/LeanIneqComp)
+[![arXiv](https://img.shields.io/badge/arXiv-2505.12680-b31b1b.svg?style=flat)](https://arxiv.org/abs/2505.12680)
+[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
+[![Email me](https://img.shields.io/badge/Contact-6fcf97?logo=gmail)](mailto:thomaszhao1998@gmail.com)
+
+</div>
+
+## Introduction
+
+We introduce Ineq-Comp, a benchmark built from elementary inequalities through systematic transformations, including variable duplication, algebraic rewriting, and multi-step composition. Although these problems remain easy for humans, we find that most provers&mdash;including Goedel, STP, and Kimina-7B&mdash;struggle significantly. DeepSeek-Prover-V2 shows relative robustness&mdash;possibly because it is trained to decompose the problems into sub-problems&mdash;but still suffers a 20\% performance drop (pass@32). Strikingly, performance remains poor for all models even when formal proofs of the constituent parts are provided in context, revealing that the source of weakness  is indeed in compositional reasoning. Our results expose a persisting gap between the generalization behavior of current AI provers and human mathematical intuition. 
+
+## Quick Start
+
+The proof of the seed problems and the evaluation scripts can be found at https://github.com/haoyuzhao123/LeanIneqComp
+
+## Ineq-Comp Benchmark
+
+We provide 5 splits: seed, type1, type2, mix, and real. For seed, type1, and type2 splits, each contains 75 problems. mix split contains 100 problems generated by Ineq-Mix, and real split contains 50 real-world inequality problems. Please refer to the github repo for more fine-grained splits.
+
+## Citation
+
+```{bibtex}
+@article{zhao2025ineq,
+  title={Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving on Inequalities},
+  author={Zhao, Haoyu and Geng, Yihan and Tang, Shange and Lin, Yong and Lyu, Bohan and Lin, Hongzhou and Jin, Chi and Arora, Sanjeev},
+  journal={arXiv preprint arXiv:2505.12680},
+  year={2025}
+}
+```