Shao-Yen Tseng

Avinash Madasu, Estelle Aflalo, Gabriela Ben Melech Stan et al.

Multi-modal retrieval has seen tremendous progress with the development of vision-language models. However, further improving these models require additional labelled data which is a huge manual effort. In this paper, we propose a framework MuMUR, that utilizes knowledge transfer from a multilingual model to boost the performance of multi-modal (image and video) retrieval. We first use state-of-the-art machine translation models to construct pseudo ground-truth multilingual visual-text pairs. We then use this data to learn a joint vision-text representation where English and non-English text queries are represented in a common embedding space based on pretrained multilingual models. We evaluate our proposed approach on a diverse set of retrieval datasets: five video retrieval datasets such as MSRVTT, MSVD, DiDeMo, Charades and MSRVTT multilingual, two image retrieval datasets such as Flickr30k and Multi30k . Experimental results demonstrate that our approach achieves state-of-the-art results on all video retrieval datasets outperforming previous models. Additionally, our framework MuMUR significantly beats other multilingual video retrieval dataset. We also observe that MuMUR exhibits strong performance on image retrieval. This demonstrates the universal ability of MuMUR to perform retrieval across all visual inputs (image and video) and text inputs (monolingual and multilingual).

Estelle Aflalo, Gabriela Ben Melech Stan, Tiep Le et al.

Large Vision Language Models (LVLMs) have achieved significant progress in integrating visual and textual inputs for multimodal reasoning. However, a recurring challenge is ensuring these models utilize visual information as effectively as linguistic content when both modalities are necessary to formulate an accurate answer. We hypothesize that hallucinations arise due to the lack of effective visual grounding in current LVLMs. Furthermore, current vision-language benchmarks are not specifically measuring the degree to which the answer require the visual input. This limitation makes it challenging to confirm that the image is truly necessary, particularly in tasks like visual question answering. In this work, we introduce FiVL, a novel method for constructing datasets designed to train LVLMs for enhanced visual grounding and also evaluate their effectiveness in achieving it. We demonstrate the value of our datasets through three approaches. First, we introduce a novel training task based on our augmented training dataset, resulting in better performance than the baseline. Second, we present benchmarks to assess the model's ability to use image as substantive evidence, rather than relying solely on linguistic priors. Finally, we identify attention heads with the strongest vision-language alignment, enabling explainability on visual-driven hallucinations. The code is available at https://github.com/IntelLabs/fivl.

Zhipeng Cai, Matthias Mueller, Reiner Birkl et al.

In the current era of generative AI breakthroughs, generating panoramic scenes from a single input image remains a key challenge. Most existing methods use diffusion-based iterative or simultaneous multi-view inpainting. However, the lack of global scene layout priors leads to subpar outputs with duplicated objects (e.g., multiple beds in a bedroom) or requires time-consuming human text inputs for each view. We propose L-MAGIC, a novel method leveraging large language models for guidance while diffusing multiple coherent views of 360 degree panoramic scenes. L-MAGIC harnesses pre-trained diffusion and language models without fine-tuning, ensuring zero-shot performance. The output quality is further enhanced by super-resolution and multi-view fusion techniques. Extensive experiments demonstrate that the resulting panoramic scenes feature better scene layouts and perspective view rendering quality compared to related works, with >70% preference in human evaluations. Combined with conditional diffusion models, L-MAGIC can accept various input modalities, including but not limited to text, depth maps, sketches, and colored scripts. Applying depth estimation further enables 3D point cloud generation and dynamic scene exploration with fluid camera motion. Code is available at https://github.com/IntelLabs/MMPano. The video presentation is available at https://youtu.be/XDMNEzH4-Ec?list=PLG9Zyvu7iBa0-a7ccNLO8LjcVRAoMn57s.

Xiao Xu, Bei Li, Chenfei Wu et al.

Two-Tower Vision-Language (VL) models have shown promising improvements on various downstream VL tasks. Although the most advanced work improves performance by building bridges between encoders, it suffers from ineffective layer-by-layer utilization of uni-modal representations and cannot flexibly exploit different levels of uni-modal semantic knowledge. In this work, we propose ManagerTower, a novel VL model architecture that gathers and combines the insights of pre-trained uni-modal experts at different levels. The managers introduced in each cross-modal layer can adaptively aggregate uni-modal semantic knowledge to facilitate more comprehensive cross-modal alignment and fusion. ManagerTower outperforms previous strong baselines both with and without Vision-Language Pre-training (VLP). With only 4M VLP data, ManagerTower achieves superior performances on various downstream VL tasks, especially 79.15% accuracy on VQAv2 Test-Std, 86.56% IR@1 and 95.64% TR@1 on Flickr30K. Code and checkpoints are available at https://github.com/LooperXX/ManagerTower.

Gabriela Ben Melech Stan, Diana Wofk, Estelle Aflalo et al.

Latent diffusion models have proven to be state-of-the-art in the creation and manipulation of visual outputs. However, as far as we know, the generation of depth maps jointly with RGB is still limited. We introduce LDM3D-VR, a suite of diffusion models targeting virtual reality development that includes LDM3D-pano and LDM3D-SR. These models enable the generation of panoramic RGBD based on textual prompts and the upscaling of low-resolution inputs to high-resolution RGBD, respectively. Our models are fine-tuned from existing pretrained models on datasets containing panoramic/high-resolution RGB images, depth maps and captions. Both models are evaluated in comparison to existing related methods.

Gabriela Ben Melech Stan, Estelle Aflalo, Raanan Yehezkel Rohekar et al.

In the rapidly evolving landscape of artificial intelligence, multi-modal large language models are emerging as a significant area of interest. These models, which combine various forms of data input, are becoming increasingly popular. However, understanding their internal mechanisms remains a complex task. Numerous advancements have been made in the field of explainability tools and mechanisms, yet there is still much to explore. In this work, we present a novel interactive application aimed towards understanding the internal mechanisms of large vision-language models. Our interface is designed to enhance the interpretability of the image patches, which are instrumental in generating an answer, and assess the efficacy of the language model in grounding its output in the image. With our application, a user can systematically investigate the model and uncover system limitations, paving the way for enhancements in system capabilities. Finally, we present a case study of how our application can aid in understanding failure mechanisms in a popular large multi-modal model: LLaVA.

Musashi Hinck, Matthew L. Olson, David Cobbley et al.

We train a suite of multimodal foundation models (MMFM) using the popular LLaVA framework with the recently released Gemma family of large language models (LLMs). Of particular interest is the 2B parameter Gemma model, which provides opportunities to construct capable small-scale MMFMs. In line with findings from other papers in this space, we test the effect of ablating three design features: pretraining the connector, utilizing a more powerful image backbone, and increasing the size of the language backbone. The resulting models, which we call LLaVA-Gemma, exhibit moderate performance on an array of evaluations, but fail to improve past the current comparably sized SOTA models. Closer analysis of performance shows mixed effects; skipping pretraining tends to reduce performance, larger vision models sometimes improve performance, and increasing language model size has inconsistent effects. We publicly release training recipes, code and weights for our models for the LLaVA-Gemma models.

Matthew Lyle Olson, Neale Ratzlaff, Musashi Hinck et al.

Although capable of generating creative text, Large Language Models (LLMs) are poor judges of what constitutes "creativity". In this work, we show that we can leverage this knowledge of how to write creatively in order to better judge what is creative. We take a mechanistic approach that extracts differences in the internal states of an LLM when prompted to respond "boringly" or "creatively" to provide a robust measure of creativity that corresponds strongly with human judgment. We also show these internal state differences can be applied to enhance the creativity of generated text at inference time.

Matthew Lyle Olson, Musashi Hinck, Neale Ratzlaff et al.

The ImageNet hierarchy provides a structured taxonomy of object categories, offering a valuable lens through which to analyze the representations learned by deep vision models. In this work, we conduct a comprehensive analysis of how vision models encode the ImageNet hierarchy, leveraging Sparse Autoencoders (SAEs) to probe their internal representations. SAEs have been widely used as an explanation tool for large language models (LLMs), where they enable the discovery of semantically meaningful features. Here, we extend their use to vision models to investigate whether learned representations align with the ontological structure defined by the ImageNet taxonomy. Our results show that SAEs uncover hierarchical relationships in model activations, revealing an implicit encoding of taxonomic structure. We analyze the consistency of these representations across different layers of the popular vision foundation model DINOv2 and provide insights into how deep vision models internalize hierarchical category information by increasing information in the class token through each layer. Our study establishes a framework for systematic hierarchical analysis of vision model representations and highlights the potential of SAEs as a tool for probing semantic structure in deep networks.

Gabriela Ben-Melech Stan, Estelle Aflalo, Man Luo et al.

Large Vision Language Models (LVLMs) have demonstrated remarkable reasoning capabilities over textual and visual inputs. However, these models remain prone to generating misinformation. Identifying and mitigating ungrounded responses is crucial for developing trustworthy AI. Traditional explainability methods such as gradient-based relevancy maps, offer insight into the decision process of models, but are often computationally expensive and unsuitable for real-time output validation. In this work, we introduce FastRM, an efficient method for predicting explainable Relevancy Maps of LVLMs. Furthermore, FastRM provides both quantitative and qualitative assessment of model confidence. Experimental results demonstrate that FastRM achieves a 99.8% reduction in computation time and a 44.4% reduction in memory footprint compared to traditional relevancy map generation. FastRM allows explainable AI to be more practical and scalable, thereby promoting its deployment in real-world applications and enabling users to more effectively evaluate the reliability of model outputs.

Matthew Lyle Olson, Musashi Hinck, Neale Ratzlaff et al.

Sparse Autoencoders (SAEs) have emerged as a popular tool for interpreting the hidden states of large language models (LLMs). By learning to reconstruct activations from a sparse bottleneck layer, SAEs discover interpretable features from the high-dimensional internal representations of LLMs. Despite their popularity with language models, SAEs remain understudied in the visual domain. In this work, we provide an extensive evaluation the representational power of SAEs for vision models using a broad range of image-based tasks. Our experimental results demonstrate that SAE features are semantically meaningful, improve out-of-distribution generalization, and enable controllable generation across three vision model architectures: vision embedding models, multi-modal LMMs and diffusion models. In vision embedding models, we find that learned SAE features can be used for OOD detection and provide evidence that they recover the ontological structure of the underlying model. For diffusion models, we demonstrate that SAEs enable semantic steering through text encoder manipulation and develop an automated pipeline for discovering human-interpretable attributes. Finally, we conduct exploratory experiments on multi-modal LLMs, finding evidence that SAE features reveal shared representations across vision and language modalities. Our study provides a foundation for SAE evaluation in vision models, highlighting their strong potential improving interpretability, generalization, and steerability in the visual domain.

Man Luo, David Cobbley, Xin Su et al.

Computer use agents (CUA) are systems that automatically interact with graphical user interfaces (GUIs) to complete tasks. CUA have made significant progress with the advent of large vision-language models (VLMs). However, these agents typically rely on cloud-based inference with substantial compute demands, raising critical privacy and scalability concerns, especially when operating on personal devices. In this work, we take a step toward privacy-preserving and resource-efficient agents by developing a lightweight vision-language model that runs entirely on local machines. To train this compact agent, we introduce an LLM-as-Judge framework that automatically evaluates and filters synthetic interaction trajectories, producing high-quality data for reinforcement learning without human annotation. Experiments on the OS-World benchmark demonstrate that our fine-tuned local model outperforms existing baselines, highlighting a promising path toward private, efficient, and generalizable GUI agents.

Neale Ratzlaff, Matthew Lyle Olson, Musashi Hinck et al.

Large Multi-Modal Models (LMMs) have demonstrated impressive capabilities as general-purpose chatbots able to engage in conversations about visual inputs. However, their responses are influenced by societal biases present in their training datasets, leading to undesirable differences in how the model responds when presented with images depicting people of different demographics. In this work, we propose a training-free debiasing framework for LMMs that intervenes on the model's representations during text generation by constructing a steering vector that reduces reference on protected attributes. Our framework introduces two complementary methods: (1) a dataset-based approach that constructs a steering vector by contrasting model activations on biased and neutral inputs, and (2) a novel optimization-based approach designed for low-resource settings, which constructs the steering vector using a single step of gradient-based perturbation without requiring additional data. Our experiments show that these interventions effectively reduce the propensity of LMMs to generate text related to protected attributes while maintaining sentiment and fluency. Furthermore, we demonstrate that debiased LMMs achieve comparable accuracy to their unmodified counterparts on downstream tasks, indicating that bias mitigation can be achieved without sacrificing model performance.

Neale Ratzlaff, Matthew Lyle Olson, Musashi Hinck et al.

Large Vision Language Models (LVLMs) such as LLaVA have demonstrated impressive capabilities as general-purpose chatbots that can engage in conversations about a provided input image. However, their responses are influenced by societal biases present in their training datasets, leading to undesirable differences in how the model responds when presented with images depicting people of different demographics. In this work, we propose a novel debiasing framework for LVLMs by directly ablating biased attributes during text generation to avoid generating text related to protected attributes, or even representing them internally. Our method requires no training and a relatively small amount of representative biased outputs (~1000 samples). Our experiments show that not only can we can minimize the propensity of LVLMs to generate text related to protected attributes, but we can even use synthetic data to inform the ablation while retaining captioning performance on real data such as COCO. Furthermore, we find the resulting generations from a debiased LVLM exhibit similar accuracy as a baseline biased model, showing that debiasing effects can be achieved without sacrificing model performance.

Gabriela Ben Melech Stan, Diana Wofk, Scottie Fox et al.

This research paper proposes a Latent Diffusion Model for 3D (LDM3D) that generates both image and depth map data from a given text prompt, allowing users to generate RGBD images from text prompts. The LDM3D model is fine-tuned on a dataset of tuples containing an RGB image, depth map and caption, and validated through extensive experiments. We also develop an application called DepthFusion, which uses the generated RGB images and depth maps to create immersive and interactive 360-degree-view experiences using TouchDesigner. This technology has the potential to transform a wide range of industries, from entertainment and gaming to architecture and design. Overall, this paper presents a significant contribution to the field of generative AI and computer vision, and showcases the potential of LDM3D and DepthFusion to revolutionize content creation and digital experiences. A short video summarizing the approach can be found at https://t.ly/tdi2.

Estelle Aflalo, Meng Du, Shao-Yen Tseng et al.

Breakthroughs in transformer-based models have revolutionized not only the NLP field, but also vision and multimodal systems. However, although visualization and interpretability tools have become available for NLP models, internal mechanisms of vision and multimodal transformers remain largely opaque. With the success of these transformers, it is increasingly critical to understand their inner workings, as unraveling these black-boxes will lead to more capable and trustworthy models. To contribute to this quest, we propose VL-InterpreT, which provides novel interactive visualizations for interpreting the attentions and hidden representations in multimodal transformers. VL-InterpreT is a task agnostic and integrated tool that (1) tracks a variety of statistics in attention heads throughout all layers for both vision and language components, (2) visualizes cross-modal and intra-modal attentions through easily readable heatmaps, and (3) plots the hidden representations of vision and language tokens as they pass through the transformer layers. In this paper, we demonstrate the functionalities of VL-InterpreT through the analysis of KD-VLP, an end-to-end pretraining vision-language multimodal transformer-based model, in the tasks of Visual Commonsense Reasoning (VCR) and WebQA, two visual question answering benchmarks. Furthermore, we also present a few interesting findings about multimodal transformer behaviors that were learned through our tool.

Vin Sachidananda, Shao-Yen Tseng, Erik Marchi et al.

Deriving multimodal representations of audio and lexical inputs is a central problem in Natural Language Understanding (NLU). In this paper, we present Contrastive Aligned Audio-Language Multirate and Multimodal Representations (CALM), an approach for learning multimodal representations using contrastive and multirate information inherent in audio and lexical inputs. The proposed model aligns acoustic and lexical information in the input embedding space of a pretrained language-only contextual embedding model. By aligning audio representations to pretrained language representations and utilizing contrastive information between acoustic inputs, CALM is able to bootstrap audio embedding competitive with existing audio representation models in only a few hours of training time. Operationally, audio spectrograms are processed using linearized patches through a Spectral Transformer (SpecTran) which is trained using a Contrastive Audio-Language Pretraining objective to align audio and language from similar queries. Subsequently, the derived acoustic and lexical tokens representations are input into a multimodal transformer to incorporate utterance level context and derive the proposed CALM representations. We show that these pretrained embeddings can subsequently be used in multimodal supervised tasks and demonstrate the benefits of the proposed pretraining steps in terms of the alignment of the two embedding spaces and the multirate nature of the pretraining. Our system shows 10-25\% improvement over existing emotion recognition systems including state-of-the-art three-modality systems under various evaluation objectives.

Yongfei Liu, Chenfei Wu, Shao-yen Tseng et al.

Self-supervised vision-and-language pretraining (VLP) aims to learn transferable multi-modal representations from large-scale image-text data and to achieve strong performances on a broad scope of vision-language tasks after finetuning. Previous mainstream VLP approaches typically adopt a two-step strategy relying on external object detectors to encode images in a multi-modal Transformer framework, which suffer from restrictive object concept space, limited image context and inefficient computation. In this paper, we propose an object-aware end-to-end VLP framework, which directly feeds image grid features from CNNs into the Transformer and learns the multi-modal representations jointly. More importantly, we propose to perform object knowledge distillation to facilitate learning cross-modal alignment at different semantic levels. To achieve that, we design two novel pretext tasks by taking object features and their semantic labels from external detectors as supervision: 1.) Object-guided masked vision modeling task focuses on enforcing object-aware representation learning in the multi-modal Transformer; 2.) Phrase-region alignment task aims to improve cross-modal alignment by utilizing the similarities between noun phrases and object labels in the linguistic space. Extensive experiments on a wide range of vision-language tasks demonstrate the efficacy of our proposed framework, and we achieve competitive or superior performances over the existing pretraining strategies.

Shao-Yen Tseng, Panayiotis Georgiou, Shrikanth Narayanan

Word embeddings such as ELMo have recently been shown to model word semantics with greater efficacy through contextualized learning on large-scale language corpora, resulting in significant improvement in state of the art across many natural language tasks. In this work we integrate acoustic information into contextualized lexical embeddings through the addition of multimodal inputs to a pretrained bidirectional language model. The language model is trained on spoken language that includes text and audio modalities. The resulting representations from this model are multimodal and contain paralinguistic information which can modify word meanings and provide affective information. We show that these multimodal embeddings can be used to improve over previous state of the art multimodal models in emotion recognition on the CMU-MOSEI dataset.

Prashanth Gurunath Shivakumar, Shao-Yen Tseng, Panayiotis Georgiou et al.

Most current language modeling techniques only exploit co-occurrence, semantic and syntactic information from the sequence of words. However, a range of information such as the state of the speaker and dynamics of the interaction might be useful. In this work we derive motivation from psycholinguistics and propose the addition of behavioral information into the context of language modeling. We propose the augmentation of language models with an additional module which analyzes the behavioral state of the current context. This behavioral information is used to gate the outputs of the language model before the final word prediction output. We show that the addition of behavioral context in language models achieves lower perplexities on behavior-rich datasets. We also confirm the validity of the proposed models on a variety of model architectures and improve on previous state-of-the-art models with generic domain Penn Treebank Corpus.

Sandeep Nallan Chakravarthula, Haoqi Li, Shao-Yen Tseng et al.

Cancer impacts the quality of life of those diagnosed as well as their spouse caregivers, in addition to potentially influencing their day-to-day behaviors. There is evidence that effective communication between spouses can improve well-being related to cancer but it is difficult to efficiently evaluate the quality of daily life interactions using manual annotation frameworks. Automated recognition of behaviors based on the interaction cues of speakers can help analyze interactions in such couples and identify behaviors which are beneficial for effective communication. In this paper, we present and detail a dataset of dyadic interactions in 85 real-life cancer-afflicted couples and a set of observational behavior codes pertaining to interpersonal communication attributes. We describe and employ neural network-based systems for classifying these behaviors based on turn-level acoustic and lexical speech patterns. Furthermore, we investigate the effect of controlling for factors such as gender, patient/caregiver role and conversation content on behavior classification. Analysis of our preliminary results indicates the challenges in this task due to the nature of the targeted behaviors and suggests that techniques incorporating contextual processing might be better suited to tackle this problem.

Shao-Yen Tseng, Brian Baucom, Panayiotis Georgiou

Unsupervised learning has been an attractive method for easily deriving meaningful data representations from vast amounts of unlabeled data. These representations, or embeddings, often yield superior results in many tasks, whether used directly or as features in subsequent training stages. However, the quality of the embeddings is highly dependent on the assumed knowledge in the unlabeled data and how the system extracts information without supervision. Domain portability is also very limited in unsupervised learning, often requiring re-training on other in-domain corpora to achieve robustness. In this work we present a multitask paradigm for unsupervised contextual learning of behavioral interactions which addresses unsupervised domain adaption. We introduce an online multitask objective into unsupervised learning and show that sentence embeddings generated through this process increases performance of affective tasks.

Shao-Yen Tseng, Juncheng Li, Yun Wang et al.

State-of-the-art audio event detection (AED) systems rely on supervised learning using strongly labeled data. However, this dependence severely limits scalability to large-scale datasets where fine resolution annotations are too expensive to obtain. In this paper, we propose a small-footprint multiple instance learning (MIL) framework for multi-class AED using weakly annotated labels. The proposed MIL framework uses audio embeddings extracted from a pre-trained convolutional neural network as input features. We show that by using audio embeddings the MIL framework can be implemented using a simple DNN with performance comparable to recurrent neural networks. We evaluate our approach by training an audio tagging system using a subset of AudioSet, which is a large collection of weakly labeled YouTube video excerpts. Combined with a late-fusion approach, we improve the F1 score of a baseline audio tagging system by 17%. We show that audio embeddings extracted by the convolutional neural networks significantly boost the performance of all MIL models. This framework reduces the model complexity of the AED system and is suitable for applications where computational resources are limited.