????? Chart topping "Sionic" models do not even begin to meet common-sense minimum expectations for posting here

#35
by spaceman7777 - opened

There is next to nothing publically available on these two 1st and 2nd place embedding models. The company website doesn't mention their existence, the README.md is next to empty, and the link provided could be malicious for all I know.

That said, "Big if True" definitely applies to these claims, but their listing here seems extremely premature, at best.

v2 is missing all files, and v1 only calls out to an API that is returning results that could very well be imaginary. Things have operated here on a common-sense level of expectations, but it wouldn't take more than a few hours to post this same collection of material, and fraudulently claim another world topping model.

Massive Text Embedding Benchmark org

Thanks for raising this! The v2 does indeed not provide any instructions so it is removed for now.
The v1 does provide instructions but indeed we don't know any details. @sionic are you planning to release details on the method etc.?

Massive Text Embedding Benchmark org

Also removed the v1 as indeed the instructions are not very clear. @sionic could you provide more details on how to exactly run the endpoint similar to the level of instructions that OpenAI provides for their endpoint?

Muennighoff changed discussion status to closed

Sign up or log in to comment