Viewing profile — mauricekleine

I built this benchmark to see how LLM performance scales with puzzle complexity on a task that’s hard to solve via pattern matching alone. It evaluates 34 models on nonograms acros…

This is the way

This is really interesting!