News for hackers. Stuff that matters.
Re:Reasoning Models Reason Well, Until They Don't by amelius on Friday October 31, 2025 @08:33AM (#45771331)
The problem is that the training data doesn't contain a lot of "I don't know".
Re:Reasoning Models Reason Well, Until They Don't by pegasus on Friday October 31, 2025 @08:49AM (#45771447)
The bigger problem is that the benchmarks / multiple-choice tests they are trained to optimize for don't distinguish between a wrong answer and "I don't know". Which is stupid and surprising. There was a thread here on HN about this recently.
Re:Reasoning Models Reason Well, Until They Don't by astrange on Friday October 31, 2025 @05:21PM (#45776836)
That's not important compared to the post-training RL, which isn't "training data".