GTA (Guess The Algorithm) Benchmark - A tool for testing AI reasoning capabilities

Modular Arithmetic Challenge

Predicting multiple algorithmic reasoning simultaneously with GNNs and LLMs

A sketch-guided multi-objective alignment framework for improving the correctness, efficiency, and algorithmic reasoning of code language models.

Reproducible experiments on executable-semantics acquisition in language models, including training-order-induced output collapse and memorization without held-out modulo generalization.

Fourier-analysis replications and extensions for grokking and algorithmic reasoning

Reverse-engineering a custom mini-transformer to mathematically prove how it learns and computes arithmetic.

Challenge your AI's algorithmic thinking with GTA Benchmark! Reverse-engineer transformations from input-output pairs. Join now! 🐙🚀

Modern PyTorch reproduction and graph extension of Neural Programmer-Interpreters

A TensorFlow CNN project analyzing the limitations of neural networks on algorithmic reasoning tasks using binary matrix datasets.

Analyze code snippets with an AI assistant to find design flaws, suggest refactoring, generate tests, and audit SOLID principles efficiently.