Understanding satire and humor is a challenging task for even current Vision-Language models. In this paper, we propose the challenging tasks of Satirical Image Detection (detecting whether an image is satirical), Understanding (generating the reason behind the image being satirical), and Completion (given one half of the image, selecting the other half from 2 given options, such that the complete image is satirical) and release a high-quality dataset YesBut, consisting of 2547 images, 1084 satirical and 1463 non-satirical, containing different artistic styles, to evaluate those tasks. Each satirical image in the dataset depicts a normal scenario, along with a conflicting scenario which is funny or ironic. Despite the success of current Vision-Language Models on multimodal tasks such as Visual QA and Image Captioning, our benchmarking experiments show that such models perform poorly on the proposed tasks on the YesBut Dataset in Zero-Shot Settings w.r.t both automated as well as human evaluation. Additionally, we release a dataset of 119 real, satirical photographs for further research. The dataset and code are available at https://github.com/abhi1nandy2/yesbut_dataset.
Multimodal Large Language Models (MLLMs) have shown promise in a broad range of vision-language tasks with their strong reasoning and generalization capabilities. However, they heavily depend on high-quality data in the Supervised Fine-Tuning (SFT) phase. The existing approaches aim to curate high-quality data via GPT-4V, but they are not scalable due to the commercial nature of GPT-4V and the simplicity of the prompts used to instruct the model. To this end, we devised the FullAnno system, which is a data engine that can generate large-scale, high-quality, and fine-grained image annotations consisting of the category and position of objects, region descriptions, text information, as well as image dense captions. This engine is characterized by its cascade annotation process, which involves multiple expert models and employs rich prompts to instruct LLMs in generating dense image captions. We re-annotated the COCO and Visual Genome datasets using our FullAnno system, tripling the number of object annotations and increasing the length of the original image captions by a factor of 15. Experiments show that the regenerated annotation can significantly enhance the capabilities of LLaVA-v1.5 on several benchmarks. The re-annotated data are available at: https://arcana-project-page.github.io
Large Multimodal Models (LMMs) have achieved significant progress by extending large language models. Building on this progress, the latest developments in LMMs demonstrate the ability to generate dense pixel-wise segmentation through the integration of segmentation models.Despite the innovations, the textual responses and segmentation masks of existing works remain at the instance level, showing limited ability to perform fine-grained understanding and segmentation even provided with detailed textual cues.To overcome this limitation, we introduce a Multi-Granularity Large Multimodal Model (MGLMM), which is capable of seamlessly adjusting the granularity of Segmentation and Captioning (SegCap) following user instructions, from panoptic SegCap to fine-grained SegCap. We name such a new task Multi-Granularity Segmentation and Captioning (MGSC). Observing the lack of a benchmark for model training and evaluation over the MGSC task, we establish a benchmark with aligned masks and captions in multi-granularity using our customized automated annotation pipeline. This benchmark comprises 10K images and more than 30K image-question pairs. We will release our dataset along with the implementation of our automated dataset annotation pipeline for further research.Besides, we propose a novel unified SegCap data format to unify heterogeneous segmentation datasets; it effectively facilitates learning to associate object concepts with visual features during multi-task training. Extensive experiments demonstrate that our MGLMM excels at tackling more than eight downstream tasks and achieves state-of-the-art performance in MGSC, GCG, image captioning, referring segmentation, multiple and empty segmentation, and reasoning segmentation tasks. The great performance and versatility of MGLMM underscore its potential impact on advancing multimodal research.
Recent studies have shown that Vision Language Large Models (VLLMs) may output content not relevant to the input images. This problem, called the hallucination phenomenon, undoubtedly degrades VLLM performance. Therefore, various anti-hallucination techniques have been proposed to make model output more reasonable and accurate. Despite their successes, from extensive tests we found that augmenting the prompt (e.g. word appending, rewriting, and spell error etc.) may change model output and make the output hallucinate again. To cure this drawback, we propose a new instruct-tuning framework called Prompt Augmentation and Caption Utilization (PACU) to boost VLLM's generation ability under the augmented prompt scenario. Concretely, on the one hand, PACU exploits existing LLMs to augment and evaluate diverse prompts automatically. The resulting high-quality prompts are utilized to enhance VLLM's ability to process different prompts. On the other hand, PACU exploits image captions to jointly work with image features as well as the prompts for response generation. When the visual feature is inaccurate, LLM can capture useful information from the image captions for response generation. Extensive experiments on hallucination evaluation and prompt-augmented datasets demonstrate that our PACU method can work well with existing schemes to effectively boost VLLM model performance. Code is available in https://github.com/zhaominyiz/PACU.
As Vision-Language Models (VLMs) advance, human-centered Assistive Technologies (ATs) for helping People with Visual Impairments (PVIs) are evolving into generalists, capable of performing multiple tasks simultaneously. However, benchmarking VLMs for ATs remains under-explored. To bridge this gap, we first create a novel AT benchmark (@Bench). Guided by a pre-design user study with PVIs, our benchmark includes the five most crucial vision-language tasks: Panoptic Segmentation, Depth Estimation, Optical Character Recognition (OCR), Image Captioning, and Visual Question Answering (VQA). Besides, we propose a novel AT model (@Model) that addresses all tasks simultaneously and can be expanded to more assistive functions for helping PVIs. Our framework exhibits outstanding performance across tasks by integrating multi-modal information, and it offers PVIs a more comprehensive assistance. Extensive experiments prove the effectiveness and generalizability of our framework.
In this paper, we describe our system under the team name Brotherhood for the English-to-Lowres Multi-Modal Translation Task. We participate in the multi-modal translation tasks for English-Hindi, English-Hausa, English-Bengali, and English-Malayalam language pairs. We present a method leveraging multi-modal Large Language Models (LLMs), specifically GPT-4o and Claude 3.5 Sonnet, to enhance cross-lingual image captioning without traditional training or fine-tuning. Our approach utilizes instruction-tuned prompting to generate rich, contextual conversations about cropped images, using their English captions as additional context. These synthetic conversations are then translated into the target languages. Finally, we employ a weighted prompting strategy, balancing the original English caption with the translated conversation to generate captions in the target language. This method achieved competitive results, scoring 37.90 BLEU on the English-Hindi Challenge Set and ranking first and second for English-Hausa on the Challenge and Evaluation Leaderboards, respectively. We conduct additional experiments on a subset of 250 images, exploring the trade-offs between BLEU scores and semantic similarity across various weighting schemes.
Today's Large Language Models (LLMs) have showcased exemplary capabilities, ranging from simple text generation to advanced image processing. Such models are currently being explored for in-vehicle services such as supporting perception tasks in Advanced Driver Assistance Systems (ADAS) or Autonomous Driving (AD) systems, given the LLMs' capabilities to process multi-modal data. However, LLMs often generate nonsensical or unfaithful information, known as ``hallucinations'': a notable issue that needs to be mitigated. In this paper, we systematically explore the adoption of SelfCheckGPT to spot hallucinations by three state-of-the-art LLMs (GPT-4o, LLaVA, and Llama3) when analysing visual automotive data from two sources: Waymo Open Dataset, from the US, and PREPER CITY dataset, from Sweden. Our results show that GPT-4o is better at generating faithful image captions than LLaVA, whereas the former demonstrated leniency in mislabeling non-hallucinated content as hallucinations compared to the latter. Furthermore, the analysis of the performance metrics revealed that the dataset type (Waymo or PREPER CITY) did not significantly affect the quality of the captions or the effectiveness of hallucination detection. However, the models showed better performance rates over images captured during daytime, compared to during dawn, dusk or night. Overall, the results show that SelfCheckGPT and its adaptation can be used to filter hallucinations in generated traffic-related image captions for state-of-the-art LLMs.
Existing vision-language understanding benchmarks largely consist of images of objects in their usual contexts. As a consequence, recent multimodal large language models can perform well with only a shallow visual understanding by relying on background language biases. Thus, strong performance on these benchmarks does not necessarily correlate with strong visual understanding. In this paper, we release JourneyBench, a comprehensive human-annotated benchmark of generated images designed to assess the model's fine-grained multimodal reasoning abilities across five tasks: complementary multimodal chain of thought, multi-image VQA, imaginary image captioning, VQA with hallucination triggers, and fine-grained retrieval with sample-specific distractors. Unlike existing benchmarks, JourneyBench explicitly requires fine-grained multimodal reasoning in unusual imaginary scenarios where language bias and holistic image gist are insufficient. We benchmark state-of-the-art models on JourneyBench and analyze performance along a number of fine-grained dimensions. Results across all five tasks show that JourneyBench is exceptionally challenging for even the best models, indicating that models' visual reasoning abilities are not as strong as they first appear. We discuss the implications of our findings and propose avenues for further research.
Vision language models (VLMs) can simultaneously reason about images and texts to tackle many tasks, from visual question answering to image captioning. This paper focuses on map parsing, a novel task that is unexplored within the VLM context and particularly useful to mobile robots. Map parsing requires understanding not only the labels but also the geometric configurations of a map, i.e., what areas are like and how they are connected. To evaluate the performance of VLMs on map parsing, we prompt VLMs with floorplan maps to generate task plans for complex indoor navigation. Our results demonstrate the remarkable capability of VLMs in map parsing, with a success rate of 0.96 in tasks requiring a sequence of nine navigation actions, e.g., approaching and going through doors. Other than intuitive observations, e.g., VLMs do better in smaller maps and simpler navigation tasks, there was a very interesting observation that its performance drops in large open areas. We provide practical suggestions to address such challenges as validated by our experimental results. Webpage: https://shorturl.at/OUkEY
Personalized outfit recommendation remains a complex challenge, demanding both fashion compatibility understanding and trend awareness. This paper presents a novel framework that harnesses the expressive power of large language models (LLMs) for this task, mitigating their "black box" and static nature through fine-tuning and direct feedback integration. We bridge the item visual-textual gap in items descriptions by employing image captioning with a Multimodal Large Language Model (MLLM). This enables the LLM to extract style and color characteristics from human-curated fashion images, forming the basis for personalized recommendations. The LLM is efficiently fine-tuned on the open-source Polyvore dataset of curated fashion images, optimizing its ability to recommend stylish outfits. A direct preference mechanism using negative examples is employed to enhance the LLM's decision-making process. This creates a self-enhancing AI feedback loop that continuously refines recommendations in line with seasonal fashion trends. Our framework is evaluated on the Polyvore dataset, demonstrating its effectiveness in two key tasks: fill-in-the-blank, and complementary item retrieval. These evaluations underline the framework's ability to generate stylish, trend-aligned outfit suggestions, continuously improving through direct feedback. The evaluation results demonstrated that our proposed framework significantly outperforms the base LLM, creating more cohesive outfits. The improved performance in these tasks underscores the proposed framework's potential to enhance the shopping experience with accurate suggestions, proving its effectiveness over the vanilla LLM based outfit generation.
Exploring the narratives conveyed by fine-art paintings is a challenge in image captioning, where the goal is to generate descriptions that not only precisely represent the visual content but also offer a in-depth interpretation of the artwork's meaning. The task is particularly complex for artwork images due to their diverse interpretations and varied aesthetic principles across different artistic schools and styles. In response to this, we present KALE Knowledge-Augmented vision-Language model for artwork Elaborations), a novel approach that enhances existing vision-language models by integrating artwork metadata as additional knowledge. KALE incorporates the metadata in two ways: firstly as direct textual input, and secondly through a multimodal heterogeneous knowledge graph. To optimize the learning of graph representations, we introduce a new cross-modal alignment loss that maximizes the similarity between the image and its corresponding metadata. Experimental results demonstrate that KALE achieves strong performance (when evaluated with CIDEr, in particular) over existing state-of-the-art work across several artwork datasets. Source code of the project is available at https://github.com/Yanbei-Jiang/Artwork-Interpretation.
We introduce Playground v3 (PGv3), our latest text-to-image model that achieves state-of-the-art (SoTA) performance across multiple testing benchmarks, excels in graphic design abilities and introduces new capabilities. Unlike traditional text-to-image generative models that rely on pre-trained language models like T5 or CLIP text encoders, our approach fully integrates Large Language Models (LLMs) with a novel structure that leverages text conditions exclusively from a decoder-only LLM. Additionally, to enhance image captioning quality-we developed an in-house captioner, capable of generating captions with varying levels of detail, enriching the diversity of text structures. We also introduce a new benchmark CapsBench to evaluate detailed image captioning performance. Experimental results demonstrate that PGv3 excels in text prompt adherence, complex reasoning, and accurate text rendering. User preference studies indicate the super-human graphic design ability of our model for common design applications, such as stickers, posters, and logo designs. Furthermore, PGv3 introduces new capabilities, including precise RGB color control and robust multilingual understanding.
The success of Vision Language Models (VLMs) on various vision-language tasks heavily relies on pre-training with large scale web-crawled datasets. However, the noisy and incomplete nature of web data makes dataset scale crucial for performance, rendering end-to-end training increasingly prohibitive. In this paper, we propose NEVLP, a noise-robust framework for efficient vision-language pre-training that requires less pre-training data. Specifically, we bridge the modality gap between a frozen image encoder and a large language model with a transformer and introduce two innovative learning strategies: noise-adaptive learning and concept-enhanced learning to mitigate the impact of noise. In noise-adaptive learning, we estimate the noise probability of each image-text pair based on the transformer's memorization effect and employ noise-adaptive regularization on image-text contrastive learning to condition cross-modal alignment. In concept-enhanced learning, we enrich incomplete text by incorporating visual concepts (objects in the image) to provide prior information about existing objects for image-text matching and image-grounded text generation, thereby mitigating text incompletion. Our framework effectively utilizes noisy web data and achieves state-of-the-art performance with less pre-training data across a wide range of vision-language tasks, including image-text retrieval, image captioning, and visual question answering.
The growth of deep learning (DL) relies heavily on huge amounts of labelled data for tasks such as natural language processing and computer vision. Specifically, in image-to-text or image-to-image pipelines, opinion (sentiment) may be inadvertently learned by a model from human-generated image captions. Additionally, learning may be affected by the variety and diversity of the provided captions. While labelling large datasets has largely relied on crowd-sourcing or data-worker pools, evaluating the quality of such training data is crucial. This study proposes an evaluation method focused on sentiment and semantic richness. That method was applied to the COCO-MS dataset, comprising approximately 150K images with segmented objects and corresponding crowd-sourced captions. We employed pre-trained models (Twitter-RoBERTa-base and BERT-base) to extract sentiment scores and variability of semantic embeddings from captions. The relation of the sentiment score and semantic variability with object categories was examined using multiple linear regression. Results indicate that while most captions were neutral, about 6% of the captions exhibited strong sentiment influenced by specific object categories. Semantic variability of within-image captions remained low and uncorrelated with object categories. Model-generated captions showed less than 1.5% of strong sentiment which was not influenced by object categories and did not correlate with the sentiment of the respective human-generated captions. This research demonstrates an approach to assess the quality of crowd- or worker-sourced captions informed by image content.
Function names can greatly aid human reverse engineers, which has spurred the development of machine learning-based approaches to predicting function names in stripped binaries. Much current work in this area now uses transformers, applying a metaphor of machine translation from code to function names. Still, function naming models face challenges in generalizing to projects unrelated to the training set. In this paper, we take a completely new approach by transferring advances in automated image captioning to the domain of binary reverse engineering, such that different parts of a binary function can be associated with parts of its name. We propose BLens, which combines multiple binary function embeddings into a new ensemble representation, aligns it with the name representation latent space via a contrastive learning approach, and generates function names with a transformer architecture tailored for function names. Our experiments demonstrate that BLens significantly outperforms the state of the art. In the usual setting of splitting per binary, we achieve an $F_1$ score of 0.79 compared to 0.70. In the cross-project setting, which emphasizes generalizability, we achieve an $F_1$ score of 0.46 compared to 0.29. Finally, in an experimental setting reducing shared components across projects, we achieve an $F_1$ score of $0.32$ compared to $0.19$.
Rapid advancements in artificial intelligence have significantly enhanced generative tasks involving music and images, employing both unimodal and multimodal approaches. This research develops a model capable of generating music that resonates with the emotions depicted in visual arts, integrating emotion labeling, image captioning, and language models to transform visual inputs into musical compositions. Addressing the scarcity of aligned art and music data, we curated the Emotion Painting Music Dataset, pairing paintings with corresponding music for effective training and evaluation. Our dual-stage framework converts images to text descriptions of emotional content and then transforms these descriptions into music, facilitating efficient learning with minimal data. Performance is evaluated using metrics such as Fr\'echet Audio Distance (FAD), Total Harmonic Distortion (THD), Inception Score (IS), and KL divergence, with audio-emotion text similarity confirmed by the pre-trained CLAP model to demonstrate high alignment between generated music and text. This synthesis tool bridges visual art and music, enhancing accessibility for the visually impaired and opening avenues in educational and therapeutic applications by providing enriched multi-sensory experiences.
Multimodal Large Language Models (MLLMs) are evaluated on various benchmarks, such as image captioning, visual question answering, and reasoning. However, many of these benchmarks include overly simple or uninformative samples, complicating the effective distinction of different MLLMs' performance. Furthermore, evaluating models across numerous benchmarks incurs a significant computational burden. To address these issues, we propose LIME (Less Is More for MLLM Evaluation), a refined and efficient benchmark curated through a semi-automated pipeline. This pipeline filters out uninformative samples and eliminates answer leakage by focusing on tasks that necessitate image-based understanding. Our experiments indicate that LIME reduces the number of samples by 76% and evaluation time by 77%, while also providing a more effective means of distinguishing the capabilities of different models. Notably, we find that traditional automatic metrics, such as CIDEr, are inadequate for assessing MLLMs' captioning performance; excluding the caption task score yields a more accurate reflection of overall model performance. All code and data are available at https://github.com/kangreen0210/LIME.
Large Vision-Language Models (LVLMs), trained on multimodal big datasets, have significantly advanced AI by excelling in vision-language tasks. However, these models remain vulnerable to adversarial attacks, particularly jailbreak attacks, which bypass safety protocols and cause the model to generate misleading or harmful responses. This vulnerability stems from both the inherent susceptibilities of LLMs and the expanded attack surface introduced by the visual modality. We propose Sim-CLIP+, a novel defense mechanism that adversarially fine-tunes the CLIP vision encoder by leveraging a Siamese architecture. This approach maximizes cosine similarity between perturbed and clean samples, facilitating resilience against adversarial manipulations. Sim-CLIP+ offers a plug-and-play solution, allowing seamless integration into existing LVLM architectures as a robust vision encoder. Unlike previous defenses, our method requires no structural modifications to the LVLM and incurs minimal computational overhead. Sim-CLIP+ demonstrates effectiveness against both gradient-based adversarial attacks and various jailbreak techniques. We evaluate Sim-CLIP+ against three distinct jailbreak attack strategies and perform clean evaluations using standard downstream datasets, including COCO for image captioning and OKVQA for visual question answering. Extensive experiments demonstrate that Sim-CLIP+ maintains high clean accuracy while substantially improving robustness against both gradient-based adversarial attacks and jailbreak techniques. Our code and robust vision encoders are available at https://github.com/speedlab-git/Robust-Encoder-against-Jailbreak-attack.git.
Aligning multiple modalities in a latent space, such as images and texts, has shown to produce powerful semantic visual representations, fueling tasks like image captioning, text-to-image generation, or image grounding. In the context of human-centric vision, albeit CLIP-like representations encode most standard human poses relatively well (such as standing or sitting), they lack sufficient acuteness to discern detailed or uncommon ones. Actually, while 3D human poses have been often associated with images (e.g. to perform pose estimation or pose-conditioned image generation), or more recently with text (e.g. for text-to-pose generation), they have seldom been paired with both. In this work, we combine 3D poses, person's pictures and textual pose descriptions to produce an enhanced 3D-, visual- and semantic-aware human pose representation. We introduce a new transformer-based model, trained in a retrieval fashion, which can take as input any combination of the aforementioned modalities. When composing modalities, it outperforms a standard multi-modal alignment retrieval model, making it possible to sort out partial information (e.g. image with the lower body occluded). We showcase the potential of such an embroidered pose representation for (1) SMPL regression from image with optional text cue; and (2) on the task of fine-grained instruction generation, which consists in generating a text that describes how to move from one 3D pose to another (as a fitness coach). Unlike prior works, our model can take any kind of input (image and/or pose) without retraining.
Although Visual-Language Models (VLMs) have shown impressive capabilities in tasks like visual question answering and image captioning, they still struggle with hallucinations. Analysis of attention distribution in these models shows that VLMs tend to processing textual tokens rather than visual tokens. This imbalance of attention distribution causes VLMs to favor textual knowledge in the case of multimodal knowledge conflicts, resulting in differences from the image information. In this paper, we propose Re-Balancing Contrastive Decoding (RBD) method, which employs textual and visual branches to recalibrate attention distribution in VLMs. Specifically, the textual branch injects image noise to stimulate the model's dependency on text, thereby reducing textual bias. Concurrently, the visual branch focuses on the selection of significant tokens, refining the attention mechanism to highlight the primary subject. This dual-branch strategy enables the RBD method to diminish textual bias while enhancing visual information. Experimental results demonstrate that our method, RBD, outperforms the existing methods by the CHAIR and POPE metrics, mitigate hallucinations without reducing the model's general capabilities.