lambench

Tech stack & purpose

λ-bench is a benchmark suite containing 120 pure lambda calculus programming problems designed to evaluate how well AI models can implement algorithms using lambda encodings of data structures. The problems span 12 categories—ranging from basic Church numeral arithmetic to complex algorithms like Sudoku solvers and FFT implementations—all written in Lamb, a minimal lambda calculus language with named top-level definitions. The project is built with TypeScript and Bun, providing a command-line evaluation harness that tests model-generated solutions against test cases by normalizing lambda terms and checking output correctness. The benchmark tracks performance across numerous AI models, with results published to a live leaderboard, and is maintained as an open-source project under the MIT license.

Languages

Contributors

• victortaelin34 commits
• 
  Victor Taelin2 commits · 6,113 followers