The overall performance in MM-NIAH is shown in Tab. 2, which is obtained by averaging the performance across the six tasks in MM-NIAH. We also provide the performance of each task in Tab. The performance for each context length range is obtained by averaging the accuracy of that context length range across different needle depths. For samples containing multiple needles, we average the depths of each needle to serve as the needle depth of this sample. A.1 More findings In addition to the findings discussed in Section 4.2, we provide more findings here. Placing questions before context does NOT improve model performance. As shown in Figure 1, all models perform poorly in understanding image needles, which we attribute to the fact that models struggle to remember the details of each image in a long multimodal document. An intuitive improvement method is placing the question before the context, which allows the model to see the options first and then read the document. However, as illustrated by the error cases (see the first row in Figure 1), this approach cause models like InternVL-1.5 to fail in following the instructions in the questions. In fact, we observe that this phenomenon holds for all MLLMs, resulting in near-zero performance. Therefore, we do not provide quantitative results but qualitatively analyzed this issue.

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.

The overall performance in MM-NIAH is shown in Tab. 2, which is obtained by averaging the performance across the six tasks in MM-NIAH. We also provide the performance of each task in Tab. The performance for each context length range is obtained by averaging the accuracy of that context length range across different needle depths. For samples containing multiple needles, we average the depths of each needle to serve as the needle depth of this sample. A.1 More findings In addition to the findings discussed in Section 4.2, we provide more findings here. Placing questions before context does NOT improve model performance. As shown in Figure 1, all models perform poorly in understanding image needles, which we attribute to the fact that models struggle to remember the details of each image in a long multimodal document. An intuitive improvement method is placing the question before the context, which allows the model to see the options first and then read the document. However, as illustrated by the error cases (see the first row in Figure 1), this approach cause models like InternVL-1.5 to fail in following the instructions in the questions. In fact, we observe that this phenomenon holds for all MLLMs, resulting in near-zero performance. Therefore, we do not provide quantitative results but qualitatively analyzed this issue.

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.

In this paper, we focus on monolithic Multimodal Large Language Models (MLLMs) that integrate visual encoding and language decoding into a single LLM. In particular, we identify that existing pre-training strategies for monolithic MLLMs often suffer from unstable optimization or catastrophic forgetting. To address this issue, our core idea is to embed a new visual parameter space into a pre-trained LLM, thereby stably learning visual knowledge from noisy data while freezing the LLM. Based on this principle, we present Mono-InternVL, a novel monolithic MLLM that seamlessly integrates a set of visual experts via a multimodal mixture-of-experts structure. Moreover, we propose an innovative pre-training strategy to maximize the visual capability of Mono-InternVL, namely Endogenous Visual Pre-training (EViP). In particular, EViP is designed as a progressive learning process for visual experts, which aims to fully exploit the visual knowledge from noisy data to high-quality data. To validate our approach, we conduct extensive experiments on 16 benchmarks. Experimental results confirm the superior performance of Mono-InternVL than existing monolithic MLLMs on 13 of 16 multimodal benchmarks, e.g., +80 points over Emu3 on OCRBench. Compared to the modular baseline, i.e., InternVL-1.5, Mono-InternVL still retains comparable multimodal performance while reducing up to 67% first token latency. Code and model are released at https://huggingface.co/OpenGVLab/Mono-InternVL-2B.

The evaluation of Long Video Understanding (LVU) performance poses an important but challenging research problem. Despite previous efforts, the existing video understanding benchmarks are severely constrained by several issues, especially the insufficient lengths of videos, a lack of diversity in video types and evaluation tasks, and the inappropriateness for evaluating LVU performances. To address the above problems, we propose a new benchmark, called MLVU (Multi-task Long Video Understanding Benchmark), for the comprehensive and in-depth evaluation of LVU. MLVU presents the following critical values: 1) The substantial and flexible extension of video lengths, which enables the benchmark to evaluate LVU performance across a wide range of durations. 2) The inclusion of various video genres, e.g., movies, surveillance footage, egocentric videos, cartoons, game videos, etc., which reflects the models' LVU performances in different scenarios. 3) The development of diversified evaluation tasks, which enables a comprehensive examination of MLLMs' key abilities in long-video understanding. The empirical study with 20 latest MLLMs reveals significant room for improvement in today's technique, as all existing methods struggle with most of the evaluation tasks and exhibit severe performance degradation when handling longer videos. Additionally, it suggests that factors such as context length, image-understanding quality, and the choice of LLM backbone can play critical roles in future advancements. We anticipate that MLVU will advance the research of long video understanding by providing a comprehensive and in-depth analysis of MLLMs.

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.