Towards Multimodal Understanding via Stable Diffusion as a Task-Aware Feature Extractor

Recent advances in multimodal large language models (MLLMs) have enabled image-based question-answering capabilities. However, a key limitation is the use of CLIP as the visual encoder; while it can capture coarse global information, it often can miss fine-grained details that are relevant to the input query. To address these shortcomings, this work studies whether pre-trained text-to-image diffusion models can serve as instruction-aware visual encoders. Through an analysis of their internal representations, we find diffusion features are both rich in semantics and can encode strong image-text alignment. Moreover, we find that we can leverage text conditioning to focus the model on regions relevant to the input question. We then investigate how to align these features with large language models and uncover a leakage phenomenon, where the LLM can inadvertently recover information from the original diffusion prompt. We analyze the causes of this leakage and propose a mitigation strategy. Based on these insights, we explore a simple fusion strategy that utilizes both CLIP and conditional diffusion features. We evaluate our approach on both general VQA and specialized MLLM benchmarks, demonstrating the promise of diffusion models for visual understanding, particularly in vision-centric tasks that require spatial and compositional reasoning. Our project page can be found https://vatsalag99.github.io/mustafar/.

4KAgent: Agentic Any Image to 4K Super-Resolution

We present 4KAgent, a unified agentic super-resolution generalist system designed to universally upscale any image to 4K resolution (and even higher, if applied iteratively). Our system can transform images from extremely low resolutions with severe degradations, for example, highly distorted inputs at 256x256, into crystal-clear, photorealistic 4K outputs. 4KAgent comprises three core components: (1) Profiling, a module that customizes the 4KAgent pipeline based on bespoke use cases; (2) A Perception Agent, which leverages vision-language models alongside image quality assessment experts to analyze the input image and make a tailored restoration plan; and (3) A Restoration Agent, which executes the plan, following a recursive execution-reflection paradigm, guided by a quality-driven mixture-of-expert policy to select the optimal output for each step. Additionally, 4KAgent embeds a specialized face restoration pipeline, significantly enhancing facial details in portrait and selfie photos. We rigorously evaluate our 4KAgent across 11 distinct task categories encompassing a total of 26 diverse benchmarks, setting new state-of-the-art on a broad spectrum of imaging domains. Our evaluations cover natural images, portrait photos, AI-generated content, satellite imagery, fluorescence microscopy, and medical imaging like fundoscopy, ultrasound, and X-ray, demonstrating superior performance in terms of both perceptual (e.g., NIQE, MUSIQ) and fidelity (e.g., PSNR) metrics. By establishing a novel agentic paradigm for low-level vision tasks, we aim to catalyze broader interest and innovation within vision-centric autonomous agents across diverse research communities. We will release all the code, models, and results at: https://4kagent.github.io.

Vision-Language-Vision Auto-Encoder: Scalable Knowledge Distillation from Diffusion Models

Building state-of-the-art Vision-Language Models (VLMs) with strong captioning capabilities typically necessitates training on billions of high-quality image-text pairs, requiring millions of GPU hours. This paper introduces the Vision-Language-Vision (VLV) auto-encoder framework, which strategically leverages key pretrained components: a vision encoder, the decoder of a Text-to-Image (T2I) diffusion model, and subsequently, a Large Language Model (LLM). Specifically, we establish an information bottleneck by regularizing the language representation space, achieved through freezing the pretrained T2I diffusion decoder. Our VLV pipeline effectively distills knowledge from the text-conditioned diffusion model using continuous embeddings, demonstrating comprehensive semantic understanding via high-quality reconstructions. Furthermore, by fine-tuning a pretrained LLM to decode the intermediate language representations into detailed descriptions, we construct a state-of-the-art (SoTA) captioner comparable to leading models like GPT-4o and Gemini 2.0 Flash. Our method demonstrates exceptional cost-efficiency and significantly reduces data requirements; by primarily utilizing single-modal images for training and maximizing the utility of existing pretrained models (image encoder, T2I diffusion model, and LLM), it circumvents the need for massive paired image-text datasets, keeping the total training expenditure under $1,000 USD.

Evaluating Attribute Confusion in Fashion Text-to-Image Generation

Despite the rapid advances in Text-to-Image (T2I) generation models, their evaluation remains challenging in domains like fashion, involving complex compositional generation. Recent automated T2I evaluation methods leverage pre-trained vision-language models to measure cross-modal alignment. However, our preliminary study reveals that they are still limited in assessing rich entity-attribute semantics, facing challenges in attribute confusion, i.e., when attributes are correctly depicted but associated to the wrong entities. To address this, we build on a Visual Question Answering (VQA) localization strategy targeting one single entity at a time across both visual and textual modalities. We propose a localized human evaluation protocol and introduce a novel automatic metric, Localized VQAScore (L-VQAScore), that combines visual localization with VQA probing both correct (reflection) and miss-localized (leakage) attribute generation. On a newly curated dataset featuring challenging compositional alignment scenarios, L-VQAScore outperforms state-of-the-art T2I evaluation methods in terms of correlation with human judgments, demonstrating its strength in capturing fine-grained entity-attribute associations. We believe L-VQAScore can be a reliable and scalable alternative to subjective evaluations.

Reading a Ruler in the Wild

Accurately converting pixel measurements into absolute real-world dimensions remains a fundamental challenge in computer vision and limits progress in key applications such as biomedicine, forensics, nutritional analysis, and e-commerce. We introduce RulerNet, a deep learning framework that robustly infers scale "in the wild" by reformulating ruler reading as a unified keypoint-detection problem and by representing the ruler with geometric-progression parameters that are invariant to perspective transformations. Unlike traditional methods that rely on handcrafted thresholds or rigid, ruler-specific pipelines, RulerNet directly localizes centimeter marks using a distortion-invariant annotation and training strategy, enabling strong generalization across diverse ruler types and imaging conditions while mitigating data scarcity. We also present a scalable synthetic-data pipeline that combines graphics-based ruler generation with ControlNet to add photorealistic context, greatly increasing training diversity and improving performance. To further enhance robustness and efficiency, we propose DeepGP, a lightweight feed-forward network that regresses geometric-progression parameters from noisy marks and eliminates iterative optimization, enabling real-time scale estimation on mobile or edge devices. Experiments show that RulerNet delivers accurate, consistent, and efficient scale estimates under challenging real-world conditions. These results underscore its utility as a generalizable measurement tool and its potential for integration with other vision components for automated, scale-aware analysis in high-impact domains. A live demo is available at https://huggingface.co/spaces/ymp5078/RulerNet-Demo.

Evaluating Large Multimodal Models for Nutrition Analysis: A Benchmark Enriched with Contextual Metadata

Large Multimodal Models (LMMs) are increasingly applied to meal images for nutrition analysis. However, existing work primarily evaluates proprietary models, such as GPT-4. This leaves the broad range of LLMs underexplored. Additionally, the influence of integrating contextual metadata and its interaction with various reasoning modifiers remains largely uncharted. This work investigates how interpreting contextual metadata derived from GPS coordinates (converted to location/venue type), timestamps (transformed into meal/day type), and the food items present can enhance LMM performance in estimating key nutritional values. These values include calories, macronutrients (protein, carbohydrates, fat), and portion sizes. We also introduce ACETADA, a new food-image dataset slated for public release. This open dataset provides nutrition information verified by the dietitian and serves as the foundation for our analysis. Our evaluation across eight LMMs (four open-weight and four closed-weight) first establishes the benefit of contextual metadata integration over straightforward prompting with images alone. We then demonstrate how this incorporation of contextual information enhances the efficacy of reasoning modifiers, such as Chain-of-Thought, Multimodal Chain-of-Thought, Scale Hint, Few-Shot, and Expert Persona. Empirical results show that integrating metadata intelligently, when applied through straightforward prompting strategies, can significantly reduce the Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) in predicted nutritional values. This work highlights the potential of context-aware LMMs for improved nutrition analysis.

Integrating Pathology Foundation Models and Spatial Transcriptomics for Cellular Decomposition from Histology Images

The rapid development of digital pathology and modern deep learning has facilitated the emergence of pathology foundation models that are expected to solve general pathology problems under various disease conditions in one unified model, with or without fine-tuning. In parallel, spatial transcriptomics has emerged as a transformative technology that enables the profiling of gene expression on hematoxylin and eosin (H&E) stained histology images. Spatial transcriptomics unlocks the unprecedented opportunity to dive into existing histology images at a more granular, cellular level. In this work, we propose a lightweight and training-efficient approach to predict cellular composition directly from H&E-stained histology images by leveraging information-enriched feature embeddings extracted from pre-trained pathology foundation models. By training a lightweight multi-layer perceptron (MLP) regressor on cell-type abundances derived via cell2location, our method efficiently distills knowledge from pathology foundation models and demonstrates the ability to accurately predict cell-type compositions from histology images, without physically performing the costly spatial transcriptomics. Our method demonstrates competitive performance compared to existing methods such as Hist2Cell, while significantly reducing computational complexity.

Deep Brain Net: An Optimized Deep Learning Model for Brain tumor Detection in MRI Images Using EfficientNetB0 and ResNet50 with Transfer Learning

In recent years, deep learning has shown great promise in the automated detection and classification of brain tumors from MRI images. However, achieving high accuracy and computational efficiency remains a challenge. In this research, we propose Deep Brain Net, a novel deep learning system designed to optimize performance in the detection of brain tumors. The model integrates the strengths of two advanced neural network architectures which are EfficientNetB0 and ResNet50, combined with transfer learning to improve generalization and reduce training time. The EfficientNetB0 architecture enhances model efficiency by utilizing mobile inverted bottleneck blocks, which incorporate depth wise separable convolutions. This design significantly reduces the number of parameters and computational cost while preserving the ability of models to learn complex feature representations. The ResNet50 architecture, pre trained on large scale datasets like ImageNet, is fine tuned for brain tumor classification. Its use of residual connections allows for training deeper networks by mitigating the vanishing gradient problem and avoiding performance degradation. The integration of these components ensures that the proposed system is both computationally efficient and highly accurate. Extensive experiments performed on publicly available MRI datasets demonstrate that Deep Brain Net consistently outperforms existing state of the art methods in terms of classification accuracy, precision, recall, and computational efficiency. The result is an accuracy of 88 percent, a weighted F1 score of 88.75 percent, and a macro AUC ROC score of 98.17 percent which demonstrates the robustness and clinical potential of Deep Brain Net in assisting radiologists with brain tumor diagnosis.

GNN-ViTCap: GNN-Enhanced Multiple Instance Learning with Vision Transformers for Whole Slide Image Classification and Captioning

Microscopic assessment of histopathology images is vital for accurate cancer diagnosis and treatment. Whole Slide Image (WSI) classification and captioning have become crucial tasks in computer-aided pathology. However, microscopic WSI face challenges such as redundant patches and unknown patch positions due to subjective pathologist captures. Moreover, generating automatic pathology captions remains a significant challenge. To address these issues, we introduce a novel GNN-ViTCap framework for classification and caption generation from histopathological microscopic images. First, a visual feature extractor generates patch embeddings. Redundant patches are then removed by dynamically clustering these embeddings using deep embedded clustering and selecting representative patches via a scalar dot attention mechanism. We build a graph by connecting each node to its nearest neighbors in the similarity matrix and apply a graph neural network to capture both local and global context. The aggregated image embeddings are projected into the language model's input space through a linear layer and combined with caption tokens to fine-tune a large language model. We validate our method on the BreakHis and PatchGastric datasets. GNN-ViTCap achieves an F1 score of 0.934 and an AUC of 0.963 for classification, along with a BLEU-4 score of 0.811 and a METEOR score of 0.569 for captioning. Experimental results demonstrate that GNN-ViTCap outperforms state of the art approaches, offering a reliable and efficient solution for microscopy based patient diagnosis.

Cross-Modality Masked Learning for Survival Prediction in ICI Treated NSCLC Patients

Accurate prognosis of non-small cell lung cancer (NSCLC) patients undergoing immunotherapy is essential for personalized treatment planning, enabling informed patient decisions, and improving both treatment outcomes and quality of life. However, the lack of large, relevant datasets and effective multi-modal feature fusion strategies pose significant challenges in this domain. To address these challenges, we present a large-scale dataset and introduce a novel framework for multi-modal feature fusion aimed at enhancing the accuracy of survival prediction. The dataset comprises 3D CT images and corresponding clinical records from NSCLC patients treated with immune checkpoint inhibitors (ICI), along with progression-free survival (PFS) and overall survival (OS) data. We further propose a cross-modality masked learning approach for medical feature fusion, consisting of two distinct branches, each tailored to its respective modality: a Slice-Depth Transformer for extracting 3D features from CT images and a graph-based Transformer for learning node features and relationships among clinical variables in tabular data. The fusion process is guided by a masked modality learning strategy, wherein the model utilizes the intact modality to reconstruct missing components. This mechanism improves the integration of modality-specific features, fostering more effective inter-modality relationships and feature interactions. Our approach demonstrates superior performance in multi-modal integration for NSCLC survival prediction, surpassing existing methods and setting a new benchmark for prognostic models in this context.