GPT-6 Astra Edited Qwen's Model Weights and Fixed a Loop, Not Its Math
GPT-6 Astra took Qwen from 7.9% to 48.1% on MATH-500 by editing 0.016% of its model weights. MIT's logs show it fixed a loop and added no new math.
In November 2026 a team at MIT CSAIL gave GPT-6 Astra a Python session over Qwen3.5-0.8B, a tiny open model, and told it to make the model better at math with no training. Astra changed 0.016% of the weights, and the score on the MATH-500 test went from 7.9% to 48.1%. An outside re-run got a similar jump.
That reads like one model teaching another algebra. MIT itself says this mostly fixed a loop, so I read the published logs to see what the edit was. They back MIT up. Qwen could already solve a lot of these problems, but it kept getting stuck repeating itself before it reached an answer. Astra got it moving again, and my read is that no new math went in.

What 7.9% looks like: one line, 1,503 times
Take problem 807: the point (0,3) in polar coordinates. Qwen gets r = 3 right away, starts worrying about tan(π/2), and then settles on one sentence.
That's the normal case. MIT published 200 of these answers in full, so I counted, and 186 hit the token limit. One of them notices, writing "Wait, I'm just repeating." It writes that 175 more times. Here are all 200, before and after the edit.
What changed in the model weights: five times past zero
So Astra went after the loop, with a head start, since the task prompt had already named repetition. It traced the looping to a direction in what layers 15 and 19 each write, then picked the 64 neurons per layer that push along it hardest. The obvious move is to take a direction like that to zero. Astra kept going. The control at the end is why I believe it.
Going past zero sounds reckless until you see that zero isn't special. Removal is just one notch on a dial, and on the same problems doubling the strength doubled the score gain. I read the loop as a pressure you push against until it stops winning.
But finishing only helps if the model already knew
On fresh practice problems the edit rescued 579 answers and lost 5. It let the wrong answers out too: answers finishing wrong went from 6 to 187, nearly all of which used to hit the wall.
So rescues make a rough ruler: of the answers that newly finish, how many are right?
Gemma 3 270M, another tiny model, is the clean test because it rarely loops. None of Astra's edits moved its math score by more than about a point.
So check the token limit before you believe the jump
My read as of November 2026 is that an edit like this frees what a model already does, and better algebra still has to be trained. That's still a real fix if you run a small model that loops. I'd just check its answers harder afterward, because a wall is obvious and a confident wrong answer isn't.
When the next "LLM made a model smarter" result comes along, the first thing I'd look up is how many answers stopped hitting the limit. If that moved with the score, it's this story again.
More from the Blog
NVIDIA MotionBricks: The 15,000 fps Isn't the News
MotionBricks' 15,000 fps is mostly a counting choice. I ran it on a laptop. The news is fast, smooth and accurate at once, enough to replace game transitions.
ChatGPT Images 2.6 GIFs Act Like Sprite Sheets: They Keep the Look, Lose Count
ChatGPT Images 2.6 GIFs behave like one sprite sheet cut into frames, so they keep the look and lose count. In 18 juggler runs, none kept three balls.