Ovis-1.6: An Open-Source Multimodal Large Language Model (MLLM) Architecture Designed to Structurally Align Visual and Textual Embeddings
Marktechpost
SEPTEMBER 29, 2024
Similarly, in the RealWorldQA benchmark, Ovis outperformed leading proprietary models such as GPT4V and Qwen-VL-Plus, scoring 2230, compared to GPT4V’s 2038. By introducing a structured visual embedding strategy, Ovis enables more effective multimodal data integration, improving performance across various tasks.
111 
Let's personalize your content