Request for correction: REINVENT4 entry in OpenSciFlow model landscape

[kkkniengLiu]

Hi REINVENT4 maintainers,

I am maintaining an early open initiative called OpenSciFlow: https://github.com/OpenSciFlow

We are building a correction-friendly map of AI-for-science models and workflow metadata. REINVENT4 is currently listed as an AI molecular design and optimization tool and a future generative-chemistry workflow candidate:

• Scientific models page: https://github.com/OpenSciFlow/awesome-ai4s-workflows/blob/main/categories/scientific-models.md
• Chemistry/materials page: https://github.com/OpenSciFlow/awesome-ai4s-workflows/blob/main/categories/chemistry-materials.md
• Model manifest backlog: https://github.com/OpenSciFlow/plugin-manifest/blob/main/docs/model-manifest-backlog.md

This is not a request for support, adoption, or integration work. I am opening this only to ask whether the entry is accurate enough before we draft a future manifest.

Specific corrections that would help:

1. Is "AI molecular design and optimization tool" a fair short classification?
2. Should the canonical link be this repository, documentation, a paper, or another project page?
3. What metadata should a workflow runner record before executing REINVENT4: scoring components, model version, input config, output molecules, random seed, filters, citations, and limitations?
4. Are there license, model-weight, or generated-molecule interpretation boundaries we should mention?
5. Which outputs are most likely to be misinterpreted by non-expert workflow users?

If Discussions are not the right place for this, please feel free to close it. A short correction comment is enough; I can make the change on the OpenSciFlow side.

[halx]

Hi,

many thanks for your interest in REINVENT4.

You are asking quite a few questions which cannot all be answered in a simple manner. For Q1 yes, for Q2 this repo.

Q3 depends very much on what you are planning to do and what the specifc workflow is or will be
Q4 license is Apache 2.0, model limitations would require a deeper discussion with a dive into the literature
Q5 Not an easy question either because it really depends on the workflow and what it exposes to the user.

Many thanks,
Hannes.