Proposal: Add a Research Paper Catalog to the MLLAM Documentation

[HetaviM29]

Hi everyone, Following the suggestion from @sadamov in neural-lam #633, I would like to start a discussion about introducing a research paper catalog in the MLLAM documentation.
Before preparing a PR, I think it would be helpful to agree on the scope and structure of the catalog.

Some discussion points:

Should the catalog include only peer-reviewed publications or also preprints (e.g., arXiv)?
What metadata should each paper include?
Title
Authors
Publication venue
Year
DOI / URL
Short description
Related topics
How should the catalog be organized?
Who should maintain and update it over time?
I'd appreciate feedback before starting the implementation.

cc @leifdenby @observingClouds

[sadamov]

Great, thanks for reviving this repo here! There has actually been quite some research built on top of neural-lam.
Some initial thoughts on your questions:

• preprint and peer-reviewed, as reviews can take >1year
• currently the repo is just a "dump" of interesting community results. I think we should reorganize this and make it nice and appealing. I don't have any strong specificson that though :)
• maintained and organized: by the mllam-community. If it becomes a lively space we can also include it in our dev-meetings. there is a slack channel called #papers where people can post their work using mllam. we could encourage that further, and link this repo here

To start scraping the web for material, neural-lam asks to cite j.oskarsson. So that is a good place to start and filter out results that used mllam tooling.

[HetaviM29]

Thanks for the feedback! I'll start by collecting the relevant papers and drafting a structured research catalog based on the suggestions here. I'll keep the initial PR focused on the catalog itself so it's easy to review, and we can iterate on additional documentation improvements later if there's interest.

[HetaviM29]

Hi , I've almost finished collecting the initial set of papers, and while organizing them I started thinking about the structure of the catalog. Since more community papers will probably be added over time, would it make sense to organize them into sections like:

• Core Neural-LAM papers
• Community papers using Neural-LAM

For each paper, I was also thinking of keeping some basic information (authors, year, venue/arXiv, short description, links to paper/code where available) so everything stays consistent. I think this would keep the catalog easy to maintain as it grows, but I'd love to hear your thoughts before I start putting together the PR.

[leifdenby]

this all sounds great @HetaviM29 ! Thanks for starting this. It might be nice to make the collection itself be a yaml-file so that people can easily add new papers.