Rio de Janeiro's "homegrown" LLM appears to be a merge of an existing model
Researchers have discovered that the 'Rio-3.5-Open-397B' model, presented as an original AI development by IplanRIO, is actually a direct element-wise merge of existing models. Evidence shows the model consistently identifies itself as a competitor's product when system prompts are removed.
Why it matters
This highlights the growing issue of 'model washing' and intellectual property theft in the open-source AI community, where existing work is rebranded as original innovation.
You must be signed in to change notification settings
The article presents technical evidence and findings from a third party without taking a political stance.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in