打破多模态视觉+语言拼接套路！ 腾讯开源Penguin-VL，直接用纯文本LLM训视觉编码器。 这项研究跳出了先有传统视觉 backbone，再接语言模型的常规路径，直接从text-only LLM初始化vision encoder。 并在2B/8B紧凑参数规模下的文档理解、长视频时序定位等复杂任务中表现出 ...

SK Telecom has unveiled a universal document interpretation technology for vision-language model (VLM) and large language model (LLM) training, based on its proprietary large language model, A.Dot X ...

Transformer-based models have rapidly spread from text to speech, vision, and other modalities. This has created challenges for the development of Neural Processing Units (NPUs). NPUs must now ...

Apple has announced its own visual language model (VLM), ' FastVLM '. Conventional VLMs have the problem of decreasing efficiency as their accuracy increases, but FastVLM maintains high accuracy while ...