Skip to content

Latest commit

 

History

History
25 lines (14 loc) · 1.83 KB

Appendix.md

File metadata and controls

25 lines (14 loc) · 1.83 KB

Below are some additional appendices added during the peer review, including:

Some examples of prompt tuning contribution to generated titles

Language Prompt-tuning Fine-tuning URL
Python How do I change variables in a function in python? How to change variables in python script https://stackoverflow.com/questions/51564669
Java Difference between spring-data-jpa and spring-boot-starter-data- Spring Boot 2 JPA https://stackoverflow.com/questions/44768112
JavaScript Setting a variable equal to another variable in JavaScript JavaScript: Why is a variable not modified when we set it to another variable? https://stackoverflow.com/questions/50840293

We can find that prompt-tuning is more informative and natural than fine-tuning in these examples. Moreover, titles generated by prompt-tuning are more similar to ground truth titles (which can be found in the URL) than those generated by fine-tuning.

Edge cases of SOTitle+ in human study

We selected and analysed some edge cases in human study.

We share the results in the Google table

User Study results

Here is our user study results. Detailed votes can be found in Google table