The TIGER-Lab/MAmmoTH2-7B-Plus is a 7 billion parameter instruction-tuned causal language model developed by TIGER-Lab, built upon the Mistral base architecture with an 8192 token context length. It is specifically optimized for enhancing reasoning abilities, particularly in mathematical tasks, by leveraging 10 million instruction-response pairs harvested from web corpora. This model demonstrates significant performance improvements on reasoning benchmarks like MATH and GSM8K, making it suitable for complex problem-solving applications.
No reviews yet. Be the first to review!
Hugging Face