NuclearAi/GPT2-137M-Reasoner-v1.0

Model Description

We're excited to share our new model, GPT2-137M-Reasoner-v1.0. This is our very first model built specifically for reasoning.

We created it using openai-community/gpt2 as a base. we used our GRPO to train it. This made the model much better at understanding and figuring things out when you ask it questions.

This first version was trained using the NuclearAi/HyperThink-Mini-50K dataset. containing over 50000 , unique User & Ai Interaction with Reasoning. We ran the training for 1 Epoch

Note : this is just the first iteration! we will soon share you more models that are trained on larger datasets and also able to reason more accurately .