Despite rapid advances, text-to-image (T2I) models still struggle to generate anatomically coherent and semantically grounded humans. We present HumanBench, a large-scale benchmark of 36K images designed to evaluate T2I models across four axes: template consistency, spatial reasoning, action understanding, and texture recognition. To quantify alignment, we propose two metrics—Agreement and Distinction—that capture both fidelity to prompts and separation from counterfactual or negated variants. HumanBench introduces a formal measure of prompt complexity based on slot counts and compositional bindings, enabling controlled analysis of model performance. Evaluating three publicly accessible models (Midjourney, Stable Diffusion, FLUX) comprehensively and three additional models on a subset, we find persistent failures, including disfigurements, species leakage, and counting errors. Prompt–image consistency declines monotonically with complexity, a trend observed across the models. A human study of 300 samples further corroborates these findings, showing a strong correlation between Agreement scores and human ratings. By unifying automated metrics, controlled prompt design, and human validation, HumanBench provides a robust foundation for auditing the compositional and anatomical capabilities of modern T2I models. Dataset is available here.

Multimodal Large Language Models (MLLMs) promise seamless integration of vision and language understanding. However, despite their strong performance, recent studies reveal that MLLMs often fail to effectively utilize visual information, frequently relying on textual cues instead. This survey provides a comprehensive analysis of the vision component in MLLMs, covering both application-level and architectural aspects. We investigate critical challenges such as weak spatial reasoning, poor fine-grained visual perception, and suboptimal fusion of visual and textual modalities. Additionally, we explore limitations in current vision encoders, benchmark inconsistencies, and their implications for downstream tasks. By synthesizing recent advancements, we highlight key research opportunities to enhance visual understanding, improve cross-modal alignment, and develop more robust and efficient MLLMs. Our observations emphasize the urgent need to elevate vision to an equal footing with language, paving the path for more reliable and perceptually aware multimodal models.

Recently, transformers have become incredibly popular in computer vision and vision-language tasks. This notable rise in their usage can be primarily attributed to the capabilities offered by attention mechanisms and the outstanding ability of transformers to adapt and apply themselves to a variety of tasks and domains. Their versatility and state-of-the-art performance have established them as indispensable tools for a wide array of applications. However, in the constantly changing landscape of machine learning, the assurance of the trustworthiness of transformers holds utmost importance. This paper conducts a thorough examination of vision-language transformers, employing three fundamental principles of responsible AI: Bias, Robustness, and Interpretability. The primary objective of this paper is to delve into the intricacies and complexities associated with the practical use of transformers, with the overarching goal of advancing our comprehension of how to enhance their reliability and accountability.

Deep learning models, such as those used for face recognition and attribute prediction, are susceptible to manipulations like adversarial noise and unintentional noise, including Gaussian and impulse noise. This paper introduces CIAI, a Class-Independent Adversarial Intent detection network built on a modified vision transformer with detection layers. CIAI employs a novel loss function that combines Maximum Mean Discrepancy and Center Loss to detect both intentional (adversarial attacks) and unintentional noise, regardless of the image class. It is trained in a multistep fashion. We also introduce the aspect of intent during detection that can act as an added layer of security. We further showcase the performance of our proposed detector on CelebA, CelebA-HQ, LFW, AgeDB, and CIFAR-10 datasets. Our detector is able to detect both intentional (like FGSM, PGD, and DeepFool) and unintentional (like Gaussian and Salt & Pepper noises) perturbations.