Spaces:
Running
on
CPU Upgrade
Why is MTEB unusable after UI upgrade?
I am a simple person. I have simple needs. I come to MTEB to get my best retrieval model.
Before the upgrade I could quickly learn that voyage-3-m-exp was the best.
Now I try the following:
- ordering by retrieval score in the table. It just doesn't work, completely messed up
- filtering by voyage. I'm gettting an empty row that is very hard to scroll horizontally. It also says 6B parameters which seems to be way off.
(This is not an ad of Voyage)
You can follow updates on GitHub https://github.com/embeddings-benchmark/mteb/issues/2224
Voyage specified 6B model size in their model card https://huggingface.co/voyageai/voyage-3-m-exp#model-information
Orderding by task type is working. Scrolling is working
Thanks for the issue @people-search-assistant .
As @Samoed says, both the filtering works; the issue is the missing scores for the voyage model. I have contacted the voyage team to ask if we can get API credits to run the multilingual benchmark.
The 6B parameters are reported by the Voyage team.