News for hackers. Stuff that matters.

A comment on Rio de Janeiro's "homegrown" LLM appears to be a merge of an existing model. Parent comment

Re:Rio de Janeiro's "homegrown" LLM appears to be a merge of... by DonsDiscountGas on Sunday June 14, 2026 @01:34PM (#48530123)
I didn't know model merging like that was possible. (Obviously possible from a pure software standpoint but I'm surprised it's effective)
[ Parent | on HN ]
Re:Rio de Janeiro's "homegrown" LLM appears to be a merge of... by bwhitty on Sunday June 14, 2026 @02:37PM (#48530939)
As another poster above linked, it’s been shown to be effective since 2022: https://arxiv.org/abs/2203.05482
[ Parent | on HN ]
Re:Rio de Janeiro's "homegrown" LLM appears to be a merge of... by nightpool on Sunday June 14, 2026 @05:27PM (#48532949)
it works because Nex N2 is also a derivative of the original base Qwen model. If it was two completely unrelated models it wouldn't work.
[ Parent | on HN ]
Re:Rio de Janeiro's "homegrown" LLM appears to be a merge of... by hypercube33 on Sunday June 14, 2026 @11:31PM (#48536272)
Even merging models with themselves as shown here in the post how they got to the top of hugging face with two gpus
[ Parent | on HN ]
Re:Rio de Janeiro's "homegrown" LLM appears to be a merge of... by baobabKoodaa on Monday June 15, 2026 @09:07AM (#48540702)
A few years back these used to be called "Frankenstein models"
[ Parent | on HN ]