adding omnis results on STS12

[perticarari]

Adding results of a test-embedding, which I titled omniembedding, on the STS12 dataset.

[perticarari]

At first, I had some misunderstandings about how to include the revision_id, but it was all fixed in the last commit.

[KennethEnevoldsen]

Hi @perticarari thanks for the PR :) We are working on a new version of the leaderboard (v2), which should give a better overview.

In that process, we are cleaning up the models, where users of the benchmark would not be able to interpret the results.

I believe this might be such a case. A user would be unable to use the model or figure out how it got the results, therefore it does not add a lot of information to the leaderboard.

Our decision has yet to be finalized so if you have strong points I would be more than happy to consider them - we want the benchmark to be as useable as possible

[perticarari]

Hi, @KennethEnevoldsen

I didn't quite understand what the issue is. What should I add / remove to get this PR approved? Is it related to the model card or model description?

[x-tabdeveloping]

The canonically accepted way to do this would probably be to first implement your model along with detailed metadata in the mteb software package and then running your model that way.
Your PR is currently missing a lot of details including references to your model on HF Hub and model details.
We are also missing information on how you implemented your evaluations.
Please submit a PR to https://github.com/embeddings-benchmark/mteb including a model implementation along with metadata, then run your model with MTEB, then submit a PR here.

[KennethEnevoldsen]

I will close this for now. Feel free to keep the discussion going, but currently it seems like this discussion has died out