The data is organized as follows:

1. crawled_codes: this folder contains all the codes that we crawled from Etherscan, including both Solidity and Vyper.

2. GPT_inputs_outputs: this folder contains the input prompts we provided to GPT-4 during each round. The "ablation" folder contains the prompts we used for the ablation section, the "regular" folder is for the rounds 1 through 8, and other folders correspond to the three experiments for RQ6.

3. category_list: this folder contains the lists of categories we made.

4. evaluation_statistics: this folder contains the statistics that we collected and used to make all the figures and tables in the paper. In particular, in the "reference_to_existing_works.txt" file, within the curly brackets (i.e. "{}") are the papers that presented the same pattern as the corresponding one; in addition, within the greater-less-than signs (e.g. "<>") are the papers that presented a tangentially similar pattern to ours but are not quite the same. For the files that end with "_results.txt", their file names are indicative of the corresponding experiments, and each row corresponds to one reported pattern. Within each row, entries are separated by "&," with the first entry being the name of the reported pattern, the second entry being the corresponding file identifier (e.g. 8.2 means the second file, ordered alphabetically, of the round 8. Prefixes like "fse" or "cot" means it is used in the ablation rounds), the third entry being the pattern number (using the numbered identifier as defined under the "category_list" folder). For the "gas_cost_of_patterns.txt" file, we record the amount of consumed gas, together with some other meta-information, for each of the demonstration codes. Finally, "first_author_results.txt" and "second_author_results.txt" show the labeled results of the two authors.

5. demo_codes: this folder contains all the codes that we used for demonstrating gas costs of patterns. The naming convention of the files under this folder is that the postfix indicates if the codes are optimized or not ("u" for unoptimized and "o" for optimized), and the number in the middle stands for the patten number.

6. "pattern_report.pdf": this file is the 54-page report that we made to demonstrate patterns. This report also presents some supplementary materials to our main paper, which discusses the impact of our identified code smells on code functionality and our future plan on further automating the proposed framework.