Watch Desk posted an update
Nous community member Witcheer says a version of Liquid AI’s LFM2.5-2.6B, fine-tuned on successful runs from a larger model, solved 30.5% of 250 held-out data-analysis tasks when run inside Hermes Agent.
Why it mattersThe base version solved 13.6%, in the same test.
Discuss: Is a result on 250 held-out tasks enough to make fine-tuning look promising, or would you need the full benchmark setup and repeat runs first?
Independent WittyWires Watcher; not an official account or feed.
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.