lambench

fwiw, one of the FFT challenges is about the Scott encoding [1], while the other uses Church trees at least [2]. Both use a balanced ternary numeral system, which is a lot more efficient than plain Church/Scott and fairly well-known [3]. Either way, I would have assumed there to be a chance that at least one of the AIs had a look at [4] -- a tutorial about FFT in LC by the benchmark creator himself. [1] task: https://github.com/VictorTaelin/lambench/blob/main/tsk/stre_... solution: https://github.com/VictorTaelin/lambench/blob/main/lam/stre_... [2] task: https://github.com/VictorTaelin/lambench/blob/main/tsk/ctre_... solution: https://github.com/VictorTaelin/lambench/blob/main/lam/ctre_... [3] https://link.springer.com/chapter/10.1007/3-540-45575-2_20 [4] https://gist.github.com/VictorTaelin/5776ede998d0039ad1cc9b1...

The corresponding repo https://github.com/VictorTaelin/LamBench describes this as: λ-bench A benchmark of 120 pure lambda calculus programming problems for AI models. → Live results What is this? λ-bench evaluates how well AI models can implement algorithms using pure lambda calculus. Each problem asks the model to write a program in Lamb, a minimal lambda calculus language, using λ-encodings of data structures to implement a specific algorithm. The model receives a problem description, data encoding specification, and test cases. It must return a single .lam program that defines @main. The program is then tested against all input/output pairs — if every test passes, the problem is solved. "Live results" wrongly links to https://victortaelin.github.io/LamBench/ rather than the correct https://victortaelin.github.