Publications by Marcella Cornia

Video-language models (VLMs) achieve strong multimodal understanding but remain prone to hallucinations, especially when reasoning about actions and temporal order. Existing mitigation strategies, such as textual filtering or random video perturbations, often fail to address the root cause: over-reliance on language priors rather than fine-grained visual dynamics. We propose a scalable framework for counterfactual video generation that synthesizes videos differing only in actions or temporal structure while preserving scene context. Our pipeline combines multimodal LLMs for action proposal and editing guidance with diffusion-based image and video models to generate semantic hard negatives at scale. Using this framework, we build CounterVid, a synthetic dataset of ~26k preference pairs targeting action recognition and temporal reasoning. We further introduce MixDPO, a unified Direct Preference Optimization approach that jointly leverages textual and visual preferences. Fine-tuning diverse VLMs with MixDPO yields consistent improvements, notably in temporal ordering, and transfers effectively to standard video hallucination benchmarks.

Large Language Models (LLMs) have revolutionized language modeling through autoregressive generation, enabling strong performance across a wide range of tasks. Recently, Diffusion Language Models (DLMs) have emerged as an alternative paradigm that generates text through iterative denoising rather than next-token prediction, allowing parallel refinement of entire sequences. While numerous diffusion-based architectures have been proposed, differences in evaluation protocols, datasets, inference budgets, and generation hyperparameters make it difficult to compare their capabilities and understand the trade-offs they offer. In this work, we present a systematic experimental analysis of modern DLMs. Specifically, we evaluate eight state-of-the-art DLMs across eight benchmarks spanning reasoning, coding, translation, knowledge, and structured problem solving, while explicitly considering both generation quality and computational efficiency. Beyond downstream evaluation, we analyze the impact of key inference-time factors, including denoising steps, context length, block size, and parallel unmasking strategies, and complement large-scale experiments with controlled comparisons of smaller models trained under identical conditions. Our analysis highlights the strengths and limitations of diffusion-based language modeling across different tasks, architectures, and inference budgets. We show that the behavior of DLMs is strongly influenced by generation-time design choices, leading to distinct trade-offs between performance and computational efficiency. Overall, our study provides practical insights into the capabilities and deployment characteristics of contemporary DLMs.

Answer grounding in visual question answering over documents remains an open challenge: most existing benchmarks lack grounding annotations or provide annotations of limited quality, and the development of answer grounding datasets is still bounded to expensive manual effort. We introduce DocAttriBench (DAB), an improved large-scale benchmark for fine-grained, element-level source attribution in Document VQA, grounding each answer to specific layout elements such as text blocks, tables, and images. To build DAB, we propose a Mask-based Perplexity-Derived Attribution method (MAPPET) that leverages document layout and language modeling to assign each answer to its most informative document element. MAPPET measures the increase in perplexity when masking candidate elements, attributing the answer to the element that contributes most to model confidence. Applying MAPPET to multiple existing Document VQA datasets yields DAB, which contains 237k documents and 296k question-answer pairs, where each answer is grounded to the corresponding layout elements. We benchmark a range of Multimodal Large Language Models (MLLMs) with grounding capabilities on DAB, evaluating answer accuracy, attribution accuracy, and overall answer quality. Our results show that while larger models tend to achieve higher answer accuracy, even the most capable models often fail to correctly localize the supporting elements. DAB thus provides a scalable benchmark for developing grounded, verifiable, and trustworthy Document VQA models.

Recent advances in Virtual Try-On (VTON) and Virtual Try-Off (VTOFF) have greatly improved photo-realistic fashion synthesis and garment reconstruction. However, existing datasets remain static, lacking instruction-driven editing for controllable and interactive fashion generation. In this work, we introduce the Dress Editing Dataset (Dress-ED), the first large-scale benchmark that unifies VTON, VTOFF, and text-guided garment editing within a single framework. Each sample in Dress-ED includes an in-shop garment image, the corresponding person image wearing the garment, their edited counterparts, and a natural-language instruction of the desired modification. Built through a fully automated multimodal pipeline that integrates MLLM-based garment understanding, diffusion-based editing, and LLM-guided verification, Dress-ED comprises over 146k verified quadruplets spanning three garment categories and seven edit types, including both appearance (e.g., color, pattern, material) and structural (e.g., sleeve length, neckline) modifications. Based on this benchmark, we further propose a unified multimodal diffusion framework that jointly reasons over linguistic instructions and visual garment cues, serving as a strong baseline for instruction-driven VTON and VTOFF. Dataset and code available at this link: https://github.com/aimagelab/Dress-ED.

Vision-and-Language Navigation (VLN) requires embodied agents to follow natural language instructions while grounding their decisions in complex visual environments. Despite recent progress, most VLN agents remain difficult to interpret, providing limited insight into why a navigation action is selected. In this work, we introduce ExplaiNavility, an interpretable VLN framework that augments navigation with explicit reasoning supervision. Our model combines a visual scene encoder with a multimodal LLM, using structured prompts to jointly condition on the instruction, the navigation history, and the candidate viewpoints available at each step. Beyond predicting the next navigable location, the model is trained to produce human-readable explanations of its decisions and descriptive low-level action labels, such as turning left, turning right, moving forward, or stopping. To supervise these outputs, we augment R2R trajectories with speaker-style action explanations generated from consecutive observations, together with action labels obtained from the simulator. By fine-tuning the embodied multimodal agent with these auxiliary reasoning objectives, ExplaiNavility encourages more transparent decision-making while preserving strong navigation ability. Experiments on the R2R benchmark show that our approach improves over prior explainable VLN methods on unseen environments, while producing interpretable rationales that make the agent's behavior easier to inspect. These results suggest that explicit reasoning supervision can benefit both navigation performance and interpretability in embodied instruction-following agents.

Sentence similarity search is a fundamental task in information retrieval, enabling applications such as search engines, question answering, and textual analysis. However, retrieval systems often struggle when training data are scarce, as is the case for low-resource languages or specialized domains such as ancient texts. To address this challenge, we propose a novel paradigm for domain-specific sentence similarity search, where the embedding space is shaped by a combination of limited real data and a large amount of synthetic data generated by Large Language Models (LLMs). Specifically, we employ LLMs to generate domain-specific sentence pairs and fine-tune a sentence embedding model, effectively distilling knowledge from the LLM to the retrieval model. We validate our method through a case study on biblical intertextuality in Latin, demonstrating that synthetic data augmentation significantly improves retrieval effectiveness in a domain with scarce annotated resources. More broadly, our approach offers a scalable and adaptable framework for enhancing retrieval in domain-specific contexts. Source code and trained models are available at https://github.com/aimagelab/biblical-retrieval-synthesis.

Despite recent advances, single-image super-resolution (SR) remains challenging, especially in real-world scenarios with complex degradations. Diffusion-based SR methods, particularly those built on Stable Diffusion, leverage strong generative priors but commonly rely on text conditioning derived from semantic captioning. Such textual descriptions provide only high-level semantics and lack the spatially aligned visual information required for faithful restoration, leading to a representation gap between abstract semantics and spatially aligned visual details. To address this limitation, we propose GramSR, a one-step diffusion-based SR framework that replaces text conditioning with dense visual features extracted from the low-resolution input using a pre-trained DINOv3 encoder. GramSR adopts a three-stage LoRA architecture, where pixel-level, semantic-level, and texture-level LoRA modules are trained sequentially. The pixel-level module focuses on degradation removal using L2 loss, the semantic-level module enhances perceptual details via LPIPS and CSD losses, and the texture-level module enforces feature correlation consistency through a Gram matrix loss computed from DINOv3 features. At inference, independent guidance scales enable flexible control over degradation removal, semantic enhancement, and texture preservation. Extensive experiments on standard SR benchmarks demonstrate that GramSR consistently outperforms existing one-step diffusion-based methods, achieving superior structural fidelity and texture realism.

Large Language Models (LLMs) are traditionally evaluated on multiple-choice question answering (MCQA) tasks using First-Token Probability (FTP), which selects the answer option whose initial token has the highest likelihood. While efficient, FTP can be fragile: models may assign high probability to unrelated tokens (misalignment) or use a valid token merely as part of a generic preamble rather than as a clear answer choice (misinterpretation), undermining the reliability of symbolic evaluation. We propose a simple solution: output prefilling, a structured natural-language prefix (e.g., 'The correct option is:') prepended to the model output. Originally explored in AI safety as an attack strategy, we repurpose prefilling to steer the model to respond with a clean, valid option, without modifying its parameters. Through extensive evaluation, we find that the FTP with prefilling strategy substantially improves accuracy, calibration, and output consistency across a broad set of LLMs and MCQA benchmarks. It outperforms standard FTP and often matches the performance of open-ended generation approaches that require full decoding and external classifiers, while being significantly more efficient. Our analysis suggests that prefilling is a simple, robust, and zero-cost method to enhance the reliability of FTP-based evaluation in multiple-choice settings.

Virtual try-on (VTON) has been widely explored for rendering garments onto person images, while its inverse task, virtual try-off (VTOFF), remains largely overlooked. VTOFF aims to recover standardized product images of garments directly from photos of clothed individuals. This capability is of great practical importance for e-commerce platforms, large-scale dataset curation, and the training of foundation models. Unlike VTON, which must handle diverse poses and styles, VTOFF naturally benefits from a consistent output format in the form of flat garment images. However, existing methods face two major limitations: (i) exclusive reliance on visual cues from a single photo often leads to ambiguity, and (ii) generated images usually suffer from loss of fine details, limiting their real-world applicability. To address these challenges, we introduce TEMU-VTOFF, a Text-Enhanced MUlti-category framework for VTOFF. Our architecture is built on a dual DiT-based backbone equipped with a multimodal attention mechanism that jointly exploits image, text, and mask information to resolve visual ambiguities and enable robust feature learning across garment categories. To explicitly mitigate detail degradation, we further design an alignment module that refines garment structures and textures, ensuring high-quality outputs. Extensive experiments on VITON-HD and Dress Code show that TEMU-VTOFF achieves new state-of-the-art performance, substantially improving both visual realism and consistency with target garments. Code and models are available at: https://temu-vtoff-page.github.io/.

Fashion illustration is a crucial medium for designers to convey their creative vision and transform design concepts into tangible representations that showcase the interplay between clothing and the human body. In the context of fashion design, computer vision techniques have the potential to enhance and streamline the design process. Departing from prior research primarily focused on virtual try-on, this paper tackles the task of multimodal-conditioned fashion image editing. Our approach aims to generate human-centric fashion images guided by multimodal prompts, including text, human body poses, garment sketches, and fabric textures. To address this problem, we propose extending latent diffusion models to incorporate these multiple modalities and modifying the structure of the denoising network, taking multimodal prompts as input. To condition the proposed architecture on fabric textures, we employ textual inversion techniques and let diverse cross-attention layers of the denoising network attend to textual and texture information, thus incorporating different granularity conditioning details. Given the lack of datasets for the task, we extend two existing fashion datasets, Dress Code and VITON-HD, with multimodal annotations. Experimental evaluations demonstrate the effectiveness of our proposed approach in terms of realism and coherence concerning the provided multimodal inputs.