News for hackers. Stuff that matters.
Re:Something is afoot in the land of Qwen by storus on Wednesday March 04, 2026 @03:40PM (#47253473)
> to decide halfway through following my detailed instructions that it would be "simpler" to just... not do what I asked

That's likely coming from the 3:1 ratio of linear to quadratic attention usage. The latest DeepSeek also suffers from it which the original R1 never exhibited.

Re:Something is afoot in the land of Qwen by nl on Thursday March 05, 2026 @01:08AM (#47258118)
There is no way you can diagnose this like that. Correlation isn't causation and much more likely is a common source of reinforcement training data.