Add eval code example

README.md

@@ -117,6 +117,7 @@ D. {choice_D}
 If the model’s output is not one of A, B, C, or D, we use GPT-4o to map the output to A–D based on the option content. 
 The final evaluation reports the absolute accuracy of the benchmark in both English and Chinese versions.
 
+Code Example: https://github.com/uni-parser/RxnBench
 
 ## 📊 Benchmark Leaderboard