# Multimodal Question Answering

docscopeOCR-7B-050425-exp is a model fine-tuned based on Qwen/Qwen2.5-VL-7B-Instruct, focusing on document-level OCR, long-context visual language understanding, and accurate image-to-text conversion of mathematical LaTeX formats.

VideoChat-R1_7B_caption is a multimodal video-text generation model based on Qwen2-VL-7B-Instruct, focusing on video content understanding and description generation.

ViCA-7B is a vision-language model fine-tuned specifically for visual-spatial reasoning in indoor video environments. Built on the LLaVA-Video-7B-Qwen2 architecture and trained using the ViCA-322K dataset, it emphasizes structured spatial annotation and instruction-based complex reasoning tasks.

DeepSeer is a vision-language model developed based on the DeepSeek-R1 model, supporting chain-of-thought reasoning and trained through dialogue templates for visual models.

VideoRefer-7B is a multimodal large language model focused on video question answering tasks, capable of understanding and analyzing spatiotemporal object relationships in videos.

LLaVA-SpaceSGG is a visual question-answering model based on LLaVA-v1.5-13b, focusing on scene graph generation tasks. It can understand image content and generate structured scene descriptions.

LongVU is a multimodal model based on Qwen2-7B, focusing on long video language understanding tasks and employing spatio-temporal adaptive compression technology.

Table LLaVA 7B is an open-source multimodal chatbot specifically designed for understanding various table images and performing diverse table-related tasks.

The Monkey Model is a large multimodal model that excels in various visual tasks by enhancing image resolution and improving text labeling methods.

InstructBLIP is the visual instruction-tuned version of BLIP-2, based on the Vicuna-13b language model, designed for vision-language tasks.

InstructBLIP is the vision-instruction-tuned version of BLIP-2, capable of generating descriptions or answers based on images and text instructions

VideoBLIP is an enhanced version of BLIP-2 capable of processing video data, using Flan T5-xl as the backbone language model.