mauricekleine

3 ms·

1. 1.▲
   by mauricekleine 9mo ago

   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 across increasing grid sizes. One thing that stood out is how dif
2. 2.▲
   Benchmarking 34 LLMs on Nonogram (Logic Puzzle) Solving(nonobench.mauricekleine.com)

   4 points by mauricekleine 9mo ago|1 comments
3. 3.▲
   by mauricekleine 2y ago

   This is the way
4. 4.▲
   by mauricekleine 4y ago

   This is really interesting!