I have put together deadline countdowns for academic venues relevant to blockchain research: https://blockchain-deadlines.github.io

If you find missing or outdated information, or fix a bug, please send a pull request on Github. The information for the venues/deadlines is maintained in a simple YAML file.

The site’s code is a fork of aideadlin.es, originally developed by Abhishek Das. Thank you! Similar forks that might be of interest to a blockchain researcher, are sec-deadlines.github.io (for security and privacy) and ds-deadlines.github.io (for distributed systems).

At first I dreaded creating the site because of the mind-numbing labor that would be required to curate the data. So I decided to give ChatGPT a shot, and it did great! After providing it with two examples, I could already copy/paste the entire call-for-papers (CFP) of a venue, and it would automatically extract the relevant information and generate the YAML entries, with a (to me) surprising accuracy. I then switched to ChatGPT Plus, where one can use plugins to enable ChatGPT to retrieve websites. I then only had to drop the venue’s website’s URL into ChatGPT, and it would retrieve the main page, and if needed also the CFP (albeit, automatically retrieving the CFP often did not work and I had to drop the URL manually), and auto-generate the YAML entries. Check out the prompt I used. Occasionally, ChatGPT started deviating from the desired output format a bit (e.g., change the order of the fields of an entry, or include the year in the venue’s title), but that was quick and easy to correct with appropriate instructions. Overall, this experience makes me optimistic that maintaining the data up-to-date can soon be done by ChatGPT almost autonomously.

Update on 2023-12-06: I just updated the deadlines according to the latest edition of each venue, which was again mostly automated with ChatGPT Plus and this prompt. Let’s see if in a few months I can perhaps feed it the entire YAML file and get it updated all at once, rather than interactively venue-by-venue.

Update on 2024-12-16: Recently, I wrote this little Python script to use the OpenAI API, and in particular its “function calling” and “structured outputs” capabilities, to update the YAML file. It works without almost any human correction (although I still double-check the updated information it produces). It is rather cumbersome to cover all edge cases in the prompt (and examples of “good” output are of little help to the model without providing all the website content that was used as “input”, which would make for an excessively long prompt), so it seems the next step will be to fine-tune the model. To collect training data for this, the Python script optionally records its “conversations” with the model.