TIGER-Lab/MAmmoTH2-8B-Plus

MAmmoTH2-8B-Plus is an 8 billion parameter language model developed by TIGER-Lab, based on the Llama-3 architecture with an 8192-token context length. It is specifically instruction-tuned using 10 million web-harvested instruction-response pairs to significantly enhance reasoning abilities, particularly in mathematical and general reasoning tasks. This model builds upon the MAmmoTH2 foundation by further training on public instruction tuning datasets, setting new performance standards in reasoning and chatbot benchmarks.