Autonomous driving visual question answering (AD-VQA) aims to answer questions related to perception, prediction, and planning based on given driving scene images, heavily relying on the model's spatial understanding capabilities. Prior works typically express spatial information through textual representations of coordinates, resulting in semantic gaps between visual coordinate representations and textual descriptions. This oversight hinders the accurate transmission of spatial information and increases the expressive burden. To address this, we propose a novel Marker-based Prompt learning framework (MPDrive), which represents spatial coordinates by concise visual markers, ensuring linguistic expressive consistency and enhancing the accuracy of both visual perception and spatial expression in AD-VQA. Specifically, we create marker images by employing a detection expert to overlay object regions with numerical labels, converting complex textual coordinate generation into straightforward text-based visual marker predictions. Moreover, we fuse original and marker images as scene-level features and integrate them with detection priors to derive instance-level features. By combining these features, we construct dual-granularity visual prompts that stimulate the LLM's spatial perception capabilities. Extensive experiments on the DriveLM and CODA-LM datasets show that MPDrive achieves state-of-the-art performance, particularly in cases requiring sophisticated spatial understanding.
Visual dialog is a vision-language task where an agent needs to answer a series of questions grounded in an image based on the understanding of the dialog history and the image. The occurrences of coreference relations in the dialog makes it a more challenging task than visual question-answering. Most previous works have focused on learning better multi-modal representations or on exploring different ways of fusing visual and language features, while the coreferences in the dialog are mainly ignored. In this paper, based on linguistic knowledge and discourse features of human dialog we propose two soft constraints that can improve the model's ability of resolving coreferences in dialog in an unsupervised way. Experimental results on the VisDial v1.0 dataset shows that our model, which integrates two novel and linguistically inspired soft constraints in a deep transformer neural architecture, obtains new state-of-the-art performance in terms of recall at 1 and other evaluation metrics compared to current existing models and this without pretraining on other vision-language datasets. Our qualitative results also demonstrate the effectiveness of the method that we propose.
Large vision-language models (LVLMs) have achieved impressive results in various visual question-answering and reasoning tasks through vision instruction tuning on specific datasets. However, there is still significant room for improvement in the alignment between visual and language modalities. Previous methods to enhance this alignment typically require external models or data, heavily depending on their capabilities and quality, which inevitably sets an upper bound on performance. In this paper, we propose SIMA, a framework that enhances visual and language modality alignment through self-improvement, eliminating the needs for external models or data. SIMA leverages prompts from existing vision instruction tuning datasets to self-generate responses and employs an in-context self-critic mechanism to select response pairs for preference tuning. The key innovation is the introduction of three vision metrics during the in-context self-critic process, which can guide the LVLM in selecting responses that enhance image comprehension. Through experiments across 14 hallucination and comprehensive benchmarks, we demonstrate that SIMA not only improves model performance across all benchmarks but also achieves superior modality alignment, outperforming previous approaches.
In real-world applications where computational resources are limited, effectively integrating visual and textual information for Visual Question Answering (VQA) presents significant challenges. This paper investigates the performance of traditional models under computational constraints, focusing on enhancing VQA performance, particularly for numerical and counting questions. We evaluate models based on Bidirectional GRU (BidGRU), GRU, Bidirectional LSTM (BidLSTM), and Convolutional Neural Networks (CNN), analyzing the impact of different vocabulary sizes, fine-tuning strategies, and embedding dimensions. Experimental results show that the BidGRU model with an embedding dimension of 300 and a vocabulary size of 3000 achieves the best overall performance without the computational overhead of larger models. Ablation studies emphasize the importance of attention mechanisms and counting information in handling complex reasoning tasks under resource limitations. Our research provides valuable insights for developing more efficient VQA models suitable for deployment in environments with limited computational capacity.
Despite the significant success of Large Vision-Language models(LVLMs), these models still suffer hallucinations when describing images, generating answers that include non-existent objects. It is reported that these models tend to over-focus on certain irrelevant image tokens that do not contain critical information for answering the question and distort the output. To address this, we propose an Instruction-Aligned Visual Attention(IAVA) approach, which identifies irrelevant tokens by comparing changes in attention weights under two different instructions. By applying contrastive decoding, we dynamically adjust the logits generated from original image tokens and irrelevant image tokens, reducing the model's over-attention to irrelevant information. The experimental results demonstrate that IAVA consistently outperforms existing decoding techniques on benchmarks such as MME, POPE, and TextVQA in mitigating object hallucinations. Our IAVA approach is available online at https://github.com/Lee-lab558/IAVA.
Video content creators need efficient tools to repurpose content, a task that often requires complex manual or automated searches. Crafting a new video from large video libraries remains a challenge. In this paper we introduce the task of Video Library Question Answering (VLQA) through an interoperable architecture that applies Retrieval Augmented Generation (RAG) to video libraries. We propose a system that uses large language models (LLMs) to generate search queries, retrieving relevant video moments indexed by speech and visual metadata. An answer generation module then integrates user queries with this metadata to produce responses with specific video timestamps. This approach shows promise in multimedia content retrieval, and AI-assisted video content creation.
Multimodal semantic understanding often has to deal with uncertainty, which means the obtained messages tend to refer to multiple targets. Such uncertainty is problematic for our interpretation, including inter- and intra-modal uncertainty. Little effort has studied the modeling of this uncertainty, particularly in pre-training on unlabeled datasets and fine-tuning in task-specific downstream datasets. In this paper, we project the representations of all modalities as probabilistic distributions via a Probability Distribution Encoder (PDE) by utilizing sequence-level interactions. Compared to the existing deterministic methods, such uncertainty modeling can convey richer multimodal semantic information and more complex relationships. Furthermore, we integrate uncertainty modeling with popular pre-training frameworks and propose suitable pre-training tasks: Distribution-based Vision-Language Contrastive learning (D-VLC), Distribution-based Masked Language Modeling (D-MLM), and Distribution-based Image-Text Matching (D-ITM). The fine-tuned models are applied to challenging downstream tasks, including image-text retrieval, visual question answering, visual reasoning, and visual entailment, and achieve state-of-the-art results.
Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities across tasks, yet they often exhibit difficulty in distinguishing task-relevant from irrelevant signals, particularly in tasks like Visual Question Answering (VQA), which can lead to susceptibility to misleading or spurious inputs. We refer to this broader limitation as the Cross-Modality Competency Problem: the model's inability to fairly evaluate all modalities. This vulnerability becomes more evident in modality-specific tasks such as image classification or pure text question answering, where models are expected to rely solely on one modality. In such tasks, spurious information from irrelevant modalities often leads to significant performance degradation. We refer to this failure as Modality Interference, which serves as a concrete and measurable instance of the cross-modality competency problem. We further design a perturbation-based causal diagnostic experiment to verify and quantify this problem. To mitigate modality interference, we propose a novel framework to fine-tune MLLMs, including perturbation-based data augmentations with both heuristic perturbations and adversarial perturbations via Projected Gradient Descent (PGD), and a consistency regularization strategy applied to model outputs with original and perturbed inputs. Experiments on multiple benchmark datasets (image-heavy, text-heavy, and VQA tasks) and multiple model families with different scales demonstrate significant improvements in robustness and cross-modality competency, indicating our method's effectiveness in boosting unimodal reasoning ability while enhancing performance on multimodal tasks.
The rise of vision-language foundation models marks an advancement in bridging the gap between human and machine capabilities in 3D scene reasoning. Existing 3D reasoning benchmarks assume real-time scene accessibility, which is impractical due to the high cost of frequent scene updates. To this end, we introduce Hypothetical 3D Reasoning, namely Hypo3D, a benchmark designed to evaluate models' ability to reason without access to real-time scene data. Models need to imagine the scene state based on a provided change description before reasoning. Hypo3D is formulated as a 3D Visual Question Answering (VQA) benchmark, comprising 7,727 context changes across 700 indoor scenes, resulting in 14,885 question-answer pairs. An anchor-based world frame is established for all scenes, ensuring consistent reference to a global frame for directional terms in context changes and QAs. Extensive experiments show that state-of-the-art foundation models struggle to reason in hypothetically changed scenes. This reveals a substantial performance gap compared to humans, particularly in scenarios involving movement changes and directional reasoning. Even when the context change is irrelevant to the question, models often incorrectly adjust their answers.
Visual knowledge bases such as Visual Genome power numerous applications in computer vision, including visual question answering and captioning, but suffer from sparse, incomplete relationships. All scene graph models to date are limited to training on a small set of visual relationships that have thousands of training labels each. Hiring human annotators is expensive, and using textual knowledge base completion methods are incompatible with visual data. In this paper, we introduce a semi-supervised method that assigns probabilistic relationship labels to a large number of unlabeled images using few labeled examples. We analyze visual relationships to suggest two types of image-agnostic features that are used to generate noisy heuristics, whose outputs are aggregated using a factor graph-based generative model. With as few as 10 labeled examples per relationship, the generative model creates enough training data to train any existing state-of-the-art scene graph model. We demonstrate that our method outperforms all baseline approaches on scene graph prediction by 5.16 recall@100 for PREDCLS. In our limited label setting, we define a complexity metric for relationships that serves as an indicator (R^2 = 0.778) for conditions under which our method succeeds over transfer learning, the de-facto approach for training with limited labels.
For vision-and-language reasoning tasks, both fully connectionist, end-to-end methods and hybrid, neuro-symbolic methods have achieved high in-distribution performance. In which out-of-distribution settings does each paradigm excel? We investigate this question on both single-image and multi-image visual question-answering through four types of generalization tests: a novel segment-combine test for multi-image queries, contrast set, compositional generalization, and cross-benchmark transfer. Vision-and-language end-to-end trained systems exhibit sizeable performance drops across all these tests. Neuro-symbolic methods suffer even more on cross-benchmark transfer from GQA to VQA, but they show smaller accuracy drops on the other generalization tests and their performance quickly improves by few-shot training. Overall, our results demonstrate the complementary benefits of these two paradigms, and emphasize the importance of using a diverse suite of generalization tests to fully characterize model robustness to distribution shift.
The inference-time resource costs of large language and vision models present a growing challenge in production deployments. We propose the use of foundation model programs, i.e., programs that can invoke foundation models with varying resource costs and performance, as an approach to this problem. Specifically, we present a method that translates a task into a program, then learns a policy for resource allocation that, on each input, selects foundation model "backends" for each program module. The policy uses smaller, cheaper backends to handle simpler subtasks, while allowing more complex subtasks to leverage larger, more capable models. We evaluate the method on two new "streaming" visual question-answering tasks in which a system answers a question on a sequence of inputs, receiving ground-truth feedback after each answer. Compared to monolithic multi-modal models, our implementation achieves up to 98% resource savings with minimal accuracy loss, demonstrating its potential for scalable and resource-efficient multi-modal inference.
Fine-grained visual perception requires Vision-Language Models (VLMs) to locate and interpret subtle evidence within complex scenes. Constructing supervision for this ability is challenging: questions must have correct, unambiguous answers that depend on actual fine-grained visual details rather than coarse cues or language priors. Verifying these properties makes manual annotation labor-intensive and automatic generation difficult for models that themselves struggle with fine-grained perception. Yet the same models can often recognize such details once the region is zoomed in. We exploit this global-to-zoom perception gap and introduce ZoomGain, a data construction pipeline that constructs questions about the full image from visual information gained only by zooming in. It verifies each answer on the zoomed view with multiple models, ensures that each question remains unambiguous in the full image and cannot be answered from language priors, and specifically generates high-quality distractors for multiple-choice questions. With this pipeline and human-only screening (no manual QA editing), we build ZoomGainBench, a 1,000-question bilingual benchmark on which the best of 28 VLMs achieves only 61.08. Failure analysis reveals that models favor a particular type of distractor when answering multiple-choice questions incorrectly. Beyond evaluation, training on just 9K ZoomGain samples substantially improves the visual perception capabilities of our 4B/9B models, enabling them to outperform several hundred-billion-parameter models across various perception benchmarks. The trained 9B model achieves SoTA performance on several benchmarks, such as 90.25 on HR-Bench 4K. These results establish global-to-zoom perception gaps as a practical basis for scaling fine-grained visual supervision for both evaluation and training. Our code is available at https://github.com/inclusionAI/ZoomGain.
Despite significant advancements in general AI, its effectiveness in the medical domain is limited by the lack of specialized medical knowledge. To address this, we formulate GMAI-VL-5.5M, a multimodal medical dataset created by converting hundreds of specialized medical datasets with various annotations into high-quality image-text pairs. This dataset offers comprehensive task coverage, diverse modalities, and rich image-text data. Building upon this dataset, we develop GMAI-VL, a general medical vision-language model, with a three-stage training strategy that enhances the integration of visual and textual information. This approach significantly improves the model's ability to process multimodal data, supporting accurate diagnoses and clinical decision-making. Experiments show that GMAI-VL achieves state-of-the-art performance across various multimodal medical tasks, including visual question answering and medical image diagnosis.
Remote sensing visual question answering (RSVQA) opens new opportunities for the use of overhead imagery by the general public, by enabling human-machine interaction with natural language. Building on the recent advances in natural language processing and computer vision, the goal of RSVQA is to answer a question formulated in natural language about a remote sensing image. Language understanding is essential to the success of the task, but has not yet been thoroughly examined in RSVQA. In particular, the problem of language biases is often overlooked in the remote sensing community, which can impact model robustness and lead to wrong conclusions about the performances of the model. Thus, the present work aims at highlighting the problem of language biases in RSVQA with a threefold analysis strategy: visual blind models, adversarial testing and dataset analysis. This analysis focuses both on model and data. Moreover, we motivate the use of more informative and complementary evaluation metrics sensitive to the issue. The gravity of language biases in RSVQA is then exposed for all of these methods with the training of models discarding the image data and the manipulation of the visual input during inference. Finally, a detailed analysis of question-answer distribution demonstrates the root of the problem in the data itself. Thanks to this analytical study, we observed that biases in remote sensing are more severe than in standard VQA, likely due to the specifics of existing remote sensing datasets for the task, e.g. geographical similarities and sparsity, as well as a simpler vocabulary and question generation strategies. While new, improved and less-biased datasets appear as a necessity for the development of the promising field of RSVQA, we demonstrate that more informed, relative evaluation metrics remain much needed to transparently communicate results of future RSVQA methods.
Neural networks often make predictions relying on the spurious correlations from the datasets rather than the intrinsic properties of the task of interest, facing sharp degradation on out-of-distribution (OOD) test data. Existing de-bias learning frameworks try to capture specific dataset bias by annotations but they fail to handle complicated OOD scenarios. Others implicitly identify the dataset bias by special design low capability biased models or losses, but they degrade when the training and testing data are from the same distribution. In this paper, we propose a General Greedy De-bias learning framework (GGD), which greedily trains the biased models and the base model. The base model is encouraged to focus on examples that are hard to solve with biased models, thus remaining robust against spurious correlations in the test stage. GGD largely improves models' OOD generalization ability on various tasks, but sometimes over-estimates the bias level and degrades on the in-distribution test. We further re-analyze the ensemble process of GGD and introduce the Curriculum Regularization inspired by curriculum learning, which achieves a good trade-off between in-distribution and out-of-distribution performance. Extensive experiments on image classification, adversarial question answering, and visual question answering demonstrate the effectiveness of our method. GGD can learn a more robust base model under the settings of both task-specific biased models with prior knowledge and self-ensemble biased model without prior knowledge.
Museums serve as vital repositories of cultural heritage and historical artifacts spanning diverse epochs, civilizations, and regions, preserving well-documented collections. Data reveal key attributes such as age, origin, material, and cultural significance. Understanding museum exhibits from their images requires reasoning beyond visual features. In this work, we facilitate such reasoning by (a) collecting and curating a large-scale dataset of 65M images and 200M question-answer pairs in the standard museum catalog format for exhibits from all around the world; (b) training large vision-language models on the collected dataset; (c) benchmarking their ability on five visual question answering tasks. The complete dataset is labeled by museum experts, ensuring the quality as well as the practical significance of the labels. We train two VLMs from different categories: the BLIP model, with vision-language aligned embeddings, but lacking the expressive power of large language models, and the LLaVA model, a powerful instruction-tuned LLM enriched with vision-language reasoning capabilities. Through exhaustive experiments, we provide several insights on the complex and fine-grained understanding of museum exhibits. In particular, we show that some questions whose answers can often be derived directly from visual features are well answered by both types of models. On the other hand, questions that require the grounding of the visual features in repositories of human knowledge are better answered by the large vision-language models, thus demonstrating their superior capacity to perform the desired reasoning. Find our dataset, benchmarks, and source code at: https://github.com/insait-institute/Museum-65
Visual grounding is a promising path toward more robust and accurate Natural Language Processing (NLP) models. Many multimodal extensions of BERT (e.g., VideoBERT, LXMERT, VL-BERT) allow a joint modeling of texts and images that lead to state-of-the-art results on multimodal tasks such as Visual Question Answering. Here, we leverage multimodal modeling for purely textual tasks (language modeling and classification) with the expectation that the multimodal pretraining provides a grounding that can improve text processing accuracy. We propose possible strategies in this respect. A first type of strategy, referred to as {\it transferred grounding} consists in applying multimodal models to text-only tasks using a placeholder to replace image input. The second one, which we call {\it associative grounding}, harnesses image retrieval to match texts with related images during both pretraining and text-only downstream tasks. We draw further distinctions into both strategies and then compare them according to their impact on language modeling and commonsense-related downstream tasks, showing improvement over text-only baselines.
Understanding data visualizations like charts and plots requires reasoning about both visual elements and numerics. Although strong in extractive questions, current chart visual question answering (chart VQA) models suffer on complex reasoning questions. In this work, we address the lack of reasoning ability by data augmentation. We leverage Large Language Models (LLMs), which have shown to have strong reasoning ability, as an automatic data annotator that generates question-answer annotations for chart images. The key innovation in our method lies in the Synthesize Step-by-Step strategy: our LLM-based data generator learns to decompose the complex question into step-by-step sub-questions (rationales), which are then used to derive the final answer using external tools, i.e. Python. This step-wise generation procedure is trained on synthetic data generated using a template-based QA generation pipeline. Experimental results highlight the significance of the proposed step-by-step generation. By training with the LLM-augmented data (LAMENDA), we significantly enhance the chart VQA models, achieving the state-of-the-art accuracy on the ChartQA and PlotQA datasets. In particular, our approach improves the accuracy of the previous state-of-the-art approach from 38% to 54% on the human-written questions in the ChartQA dataset, which needs strong reasoning. We hope our work underscores the potential of synthetic data and encourages further exploration of data augmentation using LLMs for reasoning-heavy tasks.
Recent advancements in language-model-based video understanding have been progressing at a remarkable pace, spurred by the introduction of Large Language Models (LLMs). However, the focus of prior research has been predominantly on devising a projection layer that maps video features to tokens, an approach that is both rudimentary and inefficient. In our study, we introduce a cutting-edge framework, VaQuitA, designed to refine the synergy between video and textual information. At the data level, instead of sampling frames uniformly, we implement a sampling method guided by CLIP-score rankings, which enables a more aligned selection of frames with the given question. At the feature level, we integrate a trainable Video Perceiver alongside a Visual-Query Transformer (abbreviated as VQ-Former), which bolsters the interplay between the input question and the video features. We also discover that incorporating a simple prompt, "Please be critical", into the LLM input can substantially enhance its video comprehension capabilities. Our experimental results indicate that VaQuitA consistently sets a new benchmark for zero-shot video question-answering tasks and is adept at producing high-quality, multi-turn video dialogues with users.