The paper makes it ridiculously clear what they do and don’t do. And they make the reasons why clear too. There is absolutely nothing interesting in this new article.
replyI don't understand how anyone can read the abstract and think it's "ridiculously clear" that the "Just by looking at the raw weights," doesn't apply to the Qwen experiment.
replyYou are right, if you literally just read a one paragraph description you will get less details than reading the whole post, but most of us read more than just the one paragraph. There is some actual interesting stuff in the paper, including the portion where they did edit the weights to fix the looping bug.
reply> most of us read more than just the one paragraph
replyMost of us barely read the title
Honestly, that's on them. Coddling people who are too intellectually lazy to read the entire article isn't helping anyone.
replyIt says this “All of this is possible without allowing Astra to write training loops. It uses its reasoning to figure out what each part of a neural network computes, or should compute, and changes the weights accordingly.”
replyThis seems 100% accurate to me. What is a lie here?
Nobody said they outright “lied”, most misleading statements are “100% accurate” yet are designed to mislead the reader.
replyAnd when you see the reactions in the HN comment to the original article, it's pretty clear that most people have indeed been mislead.
The article is not misleading either.
replyThe title is but then OP should talk about just the title instead
The title and the abstract are, the article is fine (but disappointing, given the excitement built up by the title + abstract)
reply>They make it sound like they improve Qwen by reading raw weights.
replyWeird, because they explicitly and clearly explain exactly what they did and didn't do.
Did you actually read the original post or just take the word of some HN comments at face value?