| ▲ | kenjackson 7 hours ago |
| The original post explains what they did and why better than this post. This post taught me nothing that wasn’t in the original. |
|
| ▲ | syllogism 6 hours ago | parent | next [-] |
| They make it sound like they improve Qwen by reading raw weights. They don't. In fact the experiments that use raw weights have very little to do with the experiments that work on larger models. |
| |
| ▲ | kenjackson 6 hours ago | parent | next [-] | | 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. | | |
| ▲ | syllogism 6 hours ago | parent [-] | | I 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. | | |
| ▲ | tedivm 5 hours ago | parent | next [-] | | You 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. | | |
| ▲ | malfist 4 hours ago | parent [-] | | > most of us read more than just the one paragraph Most of us barely read the title | | |
| ▲ | tedivm 3 hours ago | parent [-] | | Honestly, that's on them. Coddling people who are too intellectually lazy to read the entire article isn't helping anyone. |
|
| |
| ▲ | kenjackson 6 hours ago | parent | prev [-] | | It 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.” This seems 100% accurate to me. What is a lie here? | | |
| ▲ | stymaar 5 hours ago | parent [-] | | Nobody said they outright “lied”, most misleading statements are “100% accurate” yet are designed to mislead the reader. And when you see the reactions in the HN comment to the original article, it's pretty clear that most people have indeed been mislead. | | |
| ▲ | nextaccountic 5 hours ago | parent [-] | | The article is not misleading either. The title is but then OP should talk about just the title instead | | |
| ▲ | stymaar 4 hours ago | parent [-] | | The title and the abstract are, the article is fine (but disappointing, given the excitement built up by the title + abstract) |
|
|
|
|
| |
| ▲ | nater5000 4 hours ago | parent | prev [-] | | >They make it sound like they improve Qwen by reading raw weights. Weird, 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? |
|
|
| ▲ | stymaar 6 hours ago | parent | prev [-] |
| They did, except they also did write an abstract and a title that will mislead anyone who isn't going to read the full paper in details. It's fair to say that this is malpractice. |
| |
| ▲ | kenjackson 6 hours ago | parent [-] | | I just reread both and neither seem misleading to me. If anything the title is underwhelming. | | |
| ▲ | stymaar an hour ago | parent [-] | | > These results show sparks of Weight-Space Programming, which may extend beyond what we’d expect from humans or conventional algorithms. This might be a big deal for both optimization and interpretability. […] > Likewise, asked to make Qwen3.5-0.8B better at math without gradients, Astra identifies an issue in its weights causing reasoning loops that Qwen’s own model card warns about, and fixes this by natively editing 0.016% of the weights. How is that not misleading when the in the Qwen case “natively editing” refers to a mere directional ablation, a well-known technique that has been used a ton by both academics and hobbyists over the past few years? |
|
|
