MAmmoTH-VL
Explore MAmmoTH-VL, enhancing MLLMs with a 12M instruction dataset for advanced multi-modal tasks and benchmarks.
What is MAmmoTH-VL?
MAmmoTH-VL is a cutting-edge platform that enhances the reasoning capabilities of multimodal large language models through instruction tuning. It offers a massive dataset with 12 million instruction-response pairs and achieves state-of-the-art performance on benchmarks like MathVerse and MuirBench.