Update README.md

README.md

@@ -57,7 +57,7 @@ configs:
 
 - **Languages**: LLVM Intermediate Representation (LLVM IR)
 - **Size**: ~170,000 IR samples
-- **Optimization Behaviors**: >4.3 million annotations
+- **Effective Optimizations**: >4.3 million annotations
 
 ### Source
 
@@ -71,7 +71,7 @@ configs:
 IR-OptSet is suitable for the following use cases:
 
 1. **IR Understanding**: Train models to extract structural and semantic information from LLVM IR code.
-2. **Optimization Behavior Analysis**: Evaluate model ability to capture and apply real-world compiler optimizations.
+2. **IR Analysis**: Evaluate model ability to understand and analyze basic properties of LLVM IR, such as dominator trees, loop structures, and memory dependencies.
 3. **Optimized Code Generation**: Use LLMs to generate optimized IR from unoptimized input.
 
 ### Out-of-Scope Uses
@@ -94,7 +94,7 @@ Each record in IR-OptSet includes:
 - `function_name`: IR function name
 - `repo_license`: Associated license
 
-> Each `.parquet` file contains ~8,5000 examples.
+> Each `.parquet` file contains ~8,500 examples.
 
 ------
 
@@ -122,7 +122,7 @@ All transformations preserve the semantic and structural fidelity of the IR.
   - Reusable Libraries
   - Algorithms
 
-| Domain                           | Description                                                  | #Repos | #LLVM IR | #Opt. Behaviors | Avg. Eff. Opt. Steps |
+| Domain                           | Description                                                  | #Repos | #LLVM IR | #Total Eff. Opt.| Avg. Eff. Opt. Steps |
 | -------------------------------- | ------------------------------------------------------------ | ------ | -------- | --------------- | -------------------- |
 | High-Performance Computing (HPC) | Loop-intensive, memory-bound workloads; key targets for vectorization, parallelism, and memory locality optimizations. | 275    | 17,145   | 399,110         | 23.28                |
 | Machine Learning                 | Compute-bound code; used for parallel execution.             | 95     | 9,366    | 249,467         | 26.64                |
@@ -140,11 +140,11 @@ All transformations preserve the semantic and structural fidelity of the IR.
 
 ```
 @misc{iropti2025,
-  title={IROpti: Enhancing LLMs to Understand and Perform IR-level Optimizations in Compilers},
-  author={YangZi},
+  title={IR-Optset: An Optimization-Sensitive Dataset for Advancing LLM-Based IR Optimizer},
+  author={Zi Yang, Lei Qiu, Fang Lyu, Ming Zhong, Zhilei Chai, Haojie Zhou, Huimin Cui, Xiaobing Feng},
   year={2025},
-  url={https://huggingface.co/datasets/YangziResearch/IR-OptSet_Models}
+  url={https://huggingface.co/datasets/YangziResearch/IR-OptSet}
 }
 ```
 
-> We also provide an open-source toolchain for building similar datasets. If you're interested in generating your own optimization corpus, feel free to use our tools.
+> We also provide an open-source toolchain for building similar datasets. If you're interested in generating your own optimization corpus, feel free to use our tools. (https://github.com/yilingqinghan/IR-OptSet)