CogVLM: Visual Expert for Pretrained Language Models
This addresses the challenge of deep vision-language fusion for researchers and practitioners in AI, offering an open-source model with competitive performance, though it is incremental in improving existing methods.
The paper tackles the problem of integrating visual and language features in multimodal models by introducing CogVLM, a visual language foundation model that achieves state-of-the-art performance on 10 cross-modal benchmarks, such as NoCaps and GQA, and ranks second on others like VQAv2, surpassing or matching larger models like PaLI-X 55B.
We introduce CogVLM, a powerful open-source visual language foundation model. Different from the popular shallow alignment method which maps image features into the input space of language model, CogVLM bridges the gap between the frozen pretrained language model and image encoder by a trainable visual expert module in the attention and FFN layers. As a result, CogVLM enables deep fusion of vision language features without sacrificing any performance on NLP tasks. CogVLM-17B achieves state-of-the-art performance on 10 classic cross-modal benchmarks, including NoCaps, Flicker30k captioning, RefCOCO, RefCOCO+, RefCOCOg, Visual7W, GQA, ScienceQA, VizWiz VQA and TDIUC, and ranks the 2nd on VQAv2, OKVQA, TextVQA, COCO captioning, etc., surpassing or matching PaLI-X 55B. Codes and checkpoints are available at https://github.com/THUDM/CogVLM.