ZGCM-1-7B

The team of ZGCM-1-7B introduced a new model for solving mathematical problems and searching with tools. This model has a size of 7.39B parameters and achieves 97.13% accuracy on the MATH-500 test.

The ZGCM-1-7B model is also equipped with a gated sliding window attention mechanism and can handle a context size of 256K. In addition, the model was released along with recipes for its training.

To achieve high accuracy, the model was trained for 2 hours and received a rating of 10 ★ during this time.

The release of the ZGCM-1-7B model expands the capabilities of solving mathematical problems and searching with tools, providing more efficient and accurate solutions.