zhaohq/PureRL-1.5B-v6g-A-lam01-sigmoid-maskoff

Model Overview

This model, PureRL-1.5B-v6g-A-lam01-sigmoid-maskoff, is a 1.5 billion parameter language model developed by zhaohq. It is a fine-tuned version of the Qwen/Qwen2.5-Math-1.5B base model, leveraging a substantial 32768 token context length.

Key Capabilities

• Enhanced Mathematical Reasoning: The model was trained using the GRPO (Gradient-based Reward Policy Optimization) method, as introduced in the DeepSeekMath paper, specifically to improve its mathematical problem-solving abilities.
• Fine-tuned with TRL: Training was conducted using the TRL library, a framework for Transformer Reinforcement Learning.

Good For

• Applications requiring strong mathematical reasoning and logical deduction.
• Tasks where a smaller, specialized model with a long context window is beneficial for mathematical or complex problem-solving.