Multimodal Large Language Models (MLLMs) mainly fall into two architectures, each involving a trade-off between training and inference efficiency: embedding space alignment (e.g., LLaVA-1.5) is inefficient during inference, while cross-attention space alignment (e.g., Flamingo) is inefficient in training. In this paper, we compare these two architectures and identify the key factors for building efficient MLLMs. A primary difference between them lies in how attention is applied to visual tokens, particularly in their interactions with each other. To investigate whether attention among visual tokens is necessary, we propose a new self-attention mechanism, NAAViT (\textbf{N}o \textbf{A}ttention \textbf{A}mong \textbf{Vi}sual \textbf{T}okens), which eliminates this type of attention. Our pilot experiment on LLaVA-1.5 shows that attention among visual tokens is highly redundant. Based on these insights, we introduce SAISA (\textbf{S}elf-\textbf{A}ttention \textbf{I}nput \textbf{S}pace \textbf{A}lignment), a novel architecture that enhance both training and inference efficiency. SAISA directly aligns visual features with the input spaces of NAAViT self-attention blocks, reducing computational overhead in both self-attention blocks and feed-forward networks (FFNs). Using the same configuration as LLaVA-1.5, SAISA reduces inference FLOPs by 66\% and training budget by 26\%, while achieving superior performance in terms of accuracy. Comprehensive ablation studies further validate the effectiveness of SAISA across various LLMs and visual encoders. The code and model will be publicly available at https://github.com/icip-cas/SAISA.
Beyond Spurious Signals: Debiasing Multimodal Large Language Models via Counterfactual Inference and Adaptive Expert Routing
School of Computer Science, Peking University（北京大学计算机科学学院）；MOE Key Laboratory of Computational Linguistics, Peking University（北京大学教育部计算语言学重点实验室）；National Key Laboratory for Multimedia Information Processing, Peking University（北京大学国家多媒体信息处理重点实验室）
Task-Oriented Feature Compression for Multimodal Understanding via Device-Edge Co-Inference
Institute of Artificial Intelligence (TeleAI) of China Telecom（中国电信人工智能研究所）；School of Electronic and Information Engineering, Harbin Institute of Technology（哈尔滨工业大学电子与信息工程学院）；Department of Electronic and Computer Engineering, Hong Kong University of Science and Technology（香港科技大学电子与计算机工程系）；School of Software and Microelectronics, Peking University（北京大学软件与微电子学院）
Attend, Transform, or Silence: Operator-Level Visual Skipping for Efficient Multimodal LLM Inference
Tsinghua Shenzhen International Graduate School（清华大学深圳国际研究生院）；Sun Yat-sen University（中山大学）；National Supercomputing Center in Shenzhen（国家超级计算深圳中心）；Tsinghua University（清华大学）
MMTok: Multimodal Coverage Maximization for Efficient Inference of VLMs
Arizona State University（亚利桑那州立大学）；University of Washington（华盛顿大学）；University of Texas at Dallas（德克萨斯大学达拉斯分校）；Duke University（杜克大学）；Zoom Communications（Zoom通讯）
V-ITI: Mitigating Hallucinations in Multimodal Large Language Models via Visual Inference-Time Intervention
Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络安全学院）；Baidu Inc.（百度公司）
Ochanomizu University（御茶水女子大学）；RIKEN Center for Advanced Intelligence Project（理化学研究所先进智能项目中心）；Nara Institute of Science and Technology（奈良先端科学技术大学院大学）
A large amount of research about multimodal inference across text and vision has been recently developed to obtain visually grounded word and sentence representations. In this paper, we use logic-based representations as unified meaning representations for texts and images and present an unsupervised multimodal logical inference system that can effectively prove entailment relations between them. We show that by combining semantic parsing and theorem proving, the system can handle semantically complex sentences for visual-textual inference.
Shanghai University of Finance and Economics（上海财经大学）；Shanghai Key Laboratory of Data Science, Fudan University（复旦大学数据科学上海重点实验室）
We propose to solve the natural language inference problem without any supervision from the inference labels via task-agnostic multimodal pretraining. Although recent studies of multimodal self-supervised learning also represent the linguistic and visual context, their encoders for different modalities are coupled. Thus they cannot incorporate visual information when encoding plain text alone. In this paper, we propose Multimodal Aligned Contrastive Decoupled learning (MACD) network. MACD forces the decoupled text encoder to represent the visual information via contrastive learning. Therefore,…
The combination of visual and textual representations has produced excellent results in tasks such as image captioning and visual question answering, but the inference capabilities of multimodal representations are largely untested. In the case of textual representations, inference tasks such as Textual Entailment and Semantic Textual Similarity have been often used to benchmark the quality of textual representations. The long term goal of our research is to devise multimodal representation techniques that improve current inference capabilities. We thus present a novel task, Visual Semantic Te…
This paper describes a computational model of multiagent multimodal categorization that realizes emergent communication. We clarify whether the computational model can reproduce the following functions in a symbol emergence system, comprising two agents with different sensory modalities playing a naming game. (1) Function for forming a shared lexical system that comprises perceptual categories and corresponding signs, formed by agents through individual learning and semiotic communication between agents. (2) Function to improve the categorization accuracy in an agent via semiotic communication…
Bridging the Gap between Synthetic and Authentic Images for Multimodal Machine Translation
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；University of Chinese Academy of Sciences（中国科学院大学）；Beijing Language and Culture University（北京语言大学）
Understanding what sequence of steps are needed to complete a goal can help artificial intelligence systems reason about human activities. Past work in NLP has examined the task of goal-step inference for text. We introduce the visual analogue. We propose the Visual Goal-Step Inference (VGSI) task, where a model is given a textual goal and must choose which of four images represents a plausible step towards that goal. With a new dataset harvested from wikiHow consisting of 772,277 images representing human actions, we show that our task is challenging for state-of-the-art multimodal models. Mo…
Omni Demand Understanding: A Benchmark for Contextual User-Intent Inference in Multimodal Interaction
Shanghai Jiao Tong University（上海交通大学）；Shanghai Innovation Institute（上海创新研究院）；Alibaba Token Hub, Alibaba Group（阿里巴巴集团阿里巴巴Token Hub）；The Chinese University of Hong Kong（香港中文大学）；Tsinghua University（清华大学）；Hong Kong Polytechnic University（香港理工大学）；Nankai University（南开大学）；Johns Hopkins University（约翰斯·霍普金斯大学）
It is a common practice for recent works in vision language cross-modal reasoning to adopt a binary or multi-choice classification formulation taking as input a set of source image(s) and textual query. In this work, we take a sober look at such an unconditional formulation in the sense that no prior knowledge is specified with respect to the source image(s). Inspired by the designs of both visual commonsense reasoning and natural language inference tasks, we propose a new task termed Premise-based Multi-modal Reasoning(PMR) where a textual premise is the background presumption on each source …
Netaji Subhas University of Technology（尼塔吉·苏巴斯技术大学）
Fake News Detection (FND) is an essential field in natural language processing that aims to identify and check the truthfulness of major claims in a news article to decide the news veracity. FND finds its uses in preventing social, political and national damage caused due to misrepresentation of facts which may harm a certain section of society. Further, with the explosive rise in fake news dissemination over social media, including images and text, it has become imperative to identify fake news faster and more accurately. To solve this problem, this work investigates a novel multimodal stacke…
