Kukedlc/NeuralExperiment-7b-MagicCoder-v7.5

Kukedlc/NeuralExperiment-7b-MagicCoder-v7.5 is a 7 billion parameter experimental AI model fine-tuned from Kukedlc/NeuralExperiment-7b-dare-ties, designed to enhance logical reasoning, mathematical problem-solving, and programming capabilities. It was trained on specialized datasets including microsoft/orca-math-word-problems-200k, ise-uiuc/Magicoder-Evol-Instruct-110K, and sahil2801/CodeAlpaca-20k, each contributing 20,000 data points. The model focuses on addressing the 'INSTINST' bug in leaderboard models through targeted training on its latest layers, showing promising results in complex logical and mathematical puzzles.