SK Telecom has announced the open-source release of a new vision-language model (VLM) and document parsing technology for large language model (LLM) training on Hugging Face. Both innovations stem from their AI platform A.X. The A.X 4.0 VL Light model is mid-sized and trained on extensive multimodal Korean datasets, effectively interpreting complex industrial data, including tables and graphs, achieving a score of 79.4 on Korean visual benchmarks, surpassing China’s Qwen 2.5-VL32B. Additionally, SK Telecom introduced the A.X Encoder, a high-speed data processing tool optimized for LLM training, offering up to three times faster inference and double the training speed of existing models, with 149 million parameters and a natural language understanding benchmark score of 85.47, marking it at state-of-the-art levels globally. These releases contribute to SK Telecom’s initiative to establish a strong national AI foundation model, with a total of six models launched this month.
Source link
Share
Read more