Unfortunately, the modern text embedding model is a "bag of semantic tokens". We built up a small taxonomy: directional,Temporal,Casual,Comparative and Negation to demonstrate when text embedding models fails and how to mitigate the issue with fine-tuning. https://t.co/FwJWsaOfql https://t.co/f2wcQ4PvaL https://t.co/xtrHSnAnrp
— Bo (@bo_wangbo) Dec 18, 2024
from Twitter https://twitter.com/bo_wangbo
December 18, 2024 at 08:39AM
via IFTTT
No comments:
Post a Comment