I work in this field (mech interp research) and I agree, the original post is definitely misleading. Some overselling in research is normal but it's a bit much.

First, the key result has nothing to do with making Qwen better at math, it just reduces reasoning loops.

Reasoning loops are a simple problem: the model generates the </think> token (which closes reasoning) with too low probability. You can fix this with simple steering approaches where you add a vector to different components of the LLM (https://arxiv.org/abs/2503.22048, https://arxiv.org/abs/2506.18167, etc). The method Astra used for identifying the vector is the standard procedure humans use (contrastive analysis). Such steering vector creation is basically the standard "intro-level interp project" for human researchers.

If you can code and understand LLM internals, you can do this yourself in 4-8 hours. The SWE equivalent would be doing binary patching with a debugger. Cool that frontier models can do this, but not unexpected since humans can do it without issue.

Any idea why Alibaba might ship a model while aware that it has this defect when it is so easy to fix?

Is this looping tough to fix by changing the training?

This was my impression on the math result too.

Do you have thoughts on experiments one through five though? The direct one-shot reading/writing of weights seems pretty surprising.

[dead]