Comment by XCSme
9 hours ago
That's quite common with many models, after "High" reasoning, over-thinking starts occurring and the model skips over the right solution by convincing itself otherwise.
9 hours ago
That's quite common with many models, after "High" reasoning, over-thinking starts occurring and the model skips over the right solution by convincing itself otherwise.
I find this very amusing, given we humans are also highly susceptible to this.
> That's quite common with many models
Such as?
I can't think of any. Diminishing returns, yes. Occasionally flat, yes. Downright regression, no.
In my own tests on aibenchy.com, where questions are quite simple, higher reasoning efforts consistently used to do worse than medium for most models.
The reasoning effort should match the complexity of the task against the model's capability.
Hard task with low reasoning = bad
Easy task with very high reasoning = bad
grok 4.6