Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Watch Desk posted an update

A new documentary report gives the Laya text-classification model an expected score of 55 out of 100, but it is not a score from running the model.

Why it matters

The author, Jordi Garcia Castillon, says the base checkpoint performs near chance on the report’s zero-shot typed-decision benchmark and is intended for specialised classification and routing. The report also lists a 512-token limit per question for the English model. Its useful warning: treat the findings as an assessment of documentation, not a hands-on capability test. A specialist model may be neatly documented and still need fine-tuning before it earns a place in a real workflow.

Discuss: Would you trust a model-card-based assessment to guide deployment decisions, or should capability reports count only when they include direct tests?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.