Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
Jimmc414
10 months ago
|
parent
|
context
|
favorite
| on:
Bad numbers in the “gzip beats BERT” paper?
I had some difficulty reproducing results that were better than word2vec embeddings + cosine similarity. See my comment on the announcement thread here.
https://news.ycombinator.com/item?id=36706078
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: