Despite significant progress in multimodal large language models (MLLMs), their performance on complex, multi-page document comprehension remains inadequate, largely due to the lack of high-quality, document-level datasets. While current retrieval-augmented generation (RAG) methods offer partial solutions, they suffer from issues, such as fragmented retrieval contexts, multi-stage error accumulation, and extra time costs of retrieval. In this work, we present a high-quality document-level dataset, Doc-750K, designed to support in-depth understanding of multimodal documents. This dataset includes diverse document structures, extensive cross-page dependencies, and real question-answer pairs derived from the original documents. Building on the dataset, we develop a native multimodal model—Docopilot, which can accurately handle document-level dependencies without relying on RAG. Experiments demonstrate that Docopilot achieves superior coherence, accuracy, and efficiency in document understanding tasks and multi-turn interactions, setting a new baseline for document-level multimodal understanding. Data, code, and models are released at https://github.com/OpenGVLab/Docopilot.

Transformers have revolutionized computer vision and natural language processing, but their high computational complexity limits their application in high-resolution image processing and long-context analysis. This paper introduces Vision-RWKV (VRWKV), a model that builds upon the RWKV architecture from the NLP field with key modifications tailored specifically for vision tasks. Similar to the Vision Transformer (ViT), our model demonstrates robust global processing capabilities, efficiently handles sparse inputs like masked images, and can scale up to accommodate both large-scale parameters and extensive datasets. Its distinctive advantage is its reduced spatial aggregation complexity, enabling seamless processing of high-resolution images without the need for window operations. Our evaluations demonstrate that VRWKV surpasses ViT's performance in image classification and has significantly faster speeds and lower memory usage processing high-resolution inputs. In dense prediction tasks, it outperforms window-based models, maintaining comparable speeds. These results highlight VRWKV's potential as a more efficient alternative for visual perception tasks. Code and models are available at~\url{https://github.com/OpenGVLab/Vision-RWKV}.

BACKGROUND: Reliable identification of cancer subtypes is crucial for devising personalized treatment strategies. Integrating multi-omics data has proven to be an effective method for analyzing cancer subtypes. By combining molecular information across various layers, a more comprehensive understanding of biological characteristics and disease mechanisms can be achieved. RESULTS: We propose DEGCN, a novel deep learning model that integrates a three-channel Variational Autoencoder (VAE) for multi-omics dimensionality reduction and a densely connected Graph Convolutional Network (GCN) for effective subtype classification. DEGCN leverages the complementary strengths of non-linear feature extraction and graph-based relational learning, enabling accurate and robust classification of renal cancer subtypes. Experimental results demonstrate that DEGCN achieves a cross-validated classification accuracy of 97.06% ± 2.04% on renal cancer data, outperforming conventional machine learning algorithms and state-of-the-art deep learning models. Moreover, its generalization ability is validated on breast and gastric cancer datasets from TCGA, with cross-validated classification accuracies of 89.82% ± 2.29% and 88.64% ± 5.24%, respectively, indicating strong cross-cancer predictive performance. CONCLUSION: The study highlights the outstanding performance of DEGCN in heterogeneous data integration and classification accuracy, demonstrating the model's potential in cancer subtype prediction and its application in guiding clinical treatment.

With the rapid advancement of multimodal large language models (MLLMs), their evaluation has become increasingly comprehensive. However, understanding long multimodal content, as a foundational ability for real-world applications, remains underexplored. In this work, we present Needle In A Multimodal Haystack (MM-NIAH), the first benchmark specifically designed to systematically evaluate the capability of existing MLLMs to comprehend long multimodal documents. Our benchmark includes three types of evaluation tasks: multimodal retrieval, counting, and reasoning. In each task, the model is required to answer the questions according to different key information scattered throughout the given multimodal document. Evaluating the leading MLLMs on MM-NIAH, we observe that existing models still have significant room for improvement on these tasks, especially on vision-centric evaluation. We hope this work can provide a platform for further research on long multimodal document comprehension and contribute to the advancement of MLLMs. Code and benchmark are released at https://github.com/OpenGVLab/MM-NIAH.