Giordano Cicchetti

Giordano Cicchetti, Eleonora Grassucci, Danilo Comminiello

Transformer-based multimodal models rely on attention mechanisms to integrate information across heterogeneous modalities. Despite their success, existing multimodal attention formulations compute their scores through collections of pairwise dot-product interactions or by concatenating all the modalities into the keys, even when multiple modalities should be jointly involved. As a consequence, current approaches either incur quadratic complexity in the number of modalities or fail to explicitly model interactions that depend on the joint configuration of multiple representations. In this work, we introduce the Volumetric Multimodal cross-Attention (VMA), a novel cross-attention mechanism in which attention scores are defined as a function of the joint geometry of a query and multiple modality-specific keys. VMA computes the volume spanned by query and key vectors across multiple modalities, capturing joint multimodal dependencies beyond pairwise similarity, enabling native modeling of any-order modality interactions. We integrate VMA into our novel multimodal transformer architecture, named GRAMformer, explicitly designed to integrate any number of modalities. We evaluate the proposed model on multimodal learning tasks, demonstrating improved effectiveness and efficiency.

Eleonora Grassucci, Giordano Cicchetti, Emanuele Frasca et al.

In multimodal learning, CLIP has been recognized as the \textit{de facto} method for learning a shared latent space across multiple modalities, placing similar representations close to each other and moving them away from dissimilar ones. Although CLIP-based losses effectively align modalities at the semantic level, the resulting latent spaces often remain only partially shared, revealing a structural mismatch known as the modality gap. While the necessity of addressing this phenomenon remains debated, particularly given its limited impact on instance-wise tasks (e.g., retrieval), we prove that its influence is instead strongly pronounced in group-level tasks (e.g., clustering). To support this claim, we introduce a novel method designed to consistently reduce this discrepancy in two-modal settings, with a straightforward extension to the general $n$-modal case. Through our extensive evaluation, we demonstrate our novel insight: while reducing the gap provides only marginal or inconsistent improvements in traditional instance-wise tasks, it significantly enhances group-wise tasks. These findings may reshape our understanding of the modality gap, highlighting its key role in improving performance on tasks requiring semantic grouping.

Giordano Cicchetti, Eleonora Grassucci, Jihong Park et al.

In the new paradigm of semantic communication (SC), the focus is on delivering meanings behind bits by extracting semantic information from raw data. Recent advances in data-to-text models facilitate language-oriented SC, particularly for text-transformed image communication via image-to-text (I2T) encoding and text-to-image (T2I) decoding. However, although semantically aligned, the text is too coarse to precisely capture sophisticated visual features such as spatial locations, color, and texture, incurring a significant perceptual difference between intended and reconstructed images. To address this limitation, in this paper, we propose a novel language-oriented SC framework that communicates both text and a compressed image embedding and combines them using a latent diffusion model to reconstruct the intended image. Experimental results validate the potential of our approach, which transmits only 2.09\% of the original image size while achieving higher perceptual similarities in noisy communication channels compared to a baseline SC method that communicates only through text.The code is available at https://github.com/ispamm/Img2Img-SC/ .

Giordano Cicchetti, Danilo Comminiello

Real-world documents may suffer various forms of degradation, often resulting in lower accuracy in optical character recognition (OCR) systems. Therefore, a crucial preprocessing step is essential to eliminate noise while preserving text and key features of documents. In this paper, we propose NAF-DPM, a novel generative framework based on a diffusion probabilistic model (DPM) designed to restore the original quality of degraded documents. While DPMs are recognized for their high-quality generated images, they are also known for their large inference time. To mitigate this problem we provide the DPM with an efficient nonlinear activation-free (NAF) network and we employ as a sampler a fast solver of ordinary differential equations, which can converge in a few iterations. To better preserve text characters, we introduce an additional differentiable module based on convolutional recurrent neural networks, simulating the behavior of an OCR system during training. Experiments conducted on various datasets showcase the superiority of our approach, achieving state-of-the-art performance in terms of pixel-level and perceptual similarity metrics. Furthermore, the results demonstrate a notable character error reduction made by OCR systems when transcribing real-world document images enhanced by our framework. Code and pre-trained models are available at https://github.com/ispamm/NAF-DPM.

Giordano Cicchetti, Eleonora Grassucci, Luigi Sigillo et al.

Human perception integrates multiple modalities, such as vision, hearing, and language, into a unified understanding of the surrounding reality. While recent multimodal models have achieved significant progress by aligning pairs of modalities via contrastive learning, their solutions are unsuitable when scaling to multiple modalities. These models typically align each modality to a designated anchor without ensuring the alignment of all modalities with each other, leading to suboptimal performance in tasks requiring a joint understanding of multiple modalities. In this paper, we structurally rethink the pairwise conventional approach to multimodal learning and we present the novel Gramian Representation Alignment Measure (GRAM), which overcomes the above-mentioned limitations. GRAM learns and then aligns $n$ modalities directly in the higher-dimensional space in which modality embeddings lie by minimizing the Gramian volume of the $k$-dimensional parallelotope spanned by the modality vectors, ensuring the geometric alignment of all modalities simultaneously. GRAM can replace cosine similarity in any downstream method, holding for 2 to $n$ modalities and providing more meaningful alignment with respect to previous similarity measures. The novel GRAM-based contrastive loss function enhances the alignment of multimodal models in the higher-dimensional embedding space, leading to new state-of-the-art performance in downstream tasks such as video-audio-text retrieval and audio-video classification. The project page, the code, and the pretrained models are available at https://ispamm.github.io/GRAM/.

Eleonora Grassucci, Jinho Choi, Jihong Park et al.

In recent years, novel communication strategies have emerged to face the challenges that the increased number of connected devices and the higher quality of transmitted information are posing. Among them, semantic communication obtained promising results especially when combined with state-of-the-art deep generative models, such as large language or diffusion models, able to regenerate content from extremely compressed semantic information. However, most of these approaches focus on single-user scenarios processing the received content at the receiver on top of conventional communication systems. In this paper, we propose to go beyond these methods by developing a novel generative semantic communication framework tailored for multi-user scenarios. This system assigns the channel to users knowing that the lost information can be filled in with a diffusion model at the receivers. Under this innovative perspective, OFDMA systems should not aim to transmit the largest part of information, but solely the bits necessary to the generative model to semantically regenerate the missing ones. The thorough experimental evaluation shows the capabilities of the novel diffusion model and the effectiveness of the proposed framework, leading towards a GenAI-based next generation of communications.

Riccardo Fosco Gramaccioni, Christian Marinoni, Eleonora Grassucci et al.

In this work, we present FoleyGRAM, a novel approach to video-to-audio generation that emphasizes semantic conditioning through the use of aligned multimodal encoders. Building on prior advancements in video-to-audio generation, FoleyGRAM leverages the Gramian Representation Alignment Measure (GRAM) to align embeddings across video, text, and audio modalities, enabling precise semantic control over the audio generation process. The core of FoleyGRAM is a diffusion-based audio synthesis model conditioned on GRAM-aligned embeddings and waveform envelopes, ensuring both semantic richness and temporal alignment with the corresponding input video. We evaluate FoleyGRAM on the Greatest Hits dataset, a standard benchmark for video-to-audio models. Our experiments demonstrate that aligning multimodal encoders using GRAM enhances the system's ability to semantically align generated audio with video content, advancing the state of the art in video-to-audio synthesis.

Eleonora Grassucci, Giordano Cicchetti, Danilo Comminiello

In multimodal learning, CLIP has emerged as the de-facto approach for mapping different modalities into a shared latent space by bringing semantically similar representations closer while pushing apart dissimilar ones. However, CLIP-based contrastive losses exhibit unintended behaviors that negatively impact true semantic alignment, leading to sparse and fragmented latent spaces. This phenomenon, known as the modality gap, has been partially mitigated for standard text and image pairs but remains unknown and unresolved in more complex multimodal settings, such as the medical domain. In this work, we study this phenomenon in the latter case, revealing that the modality gap is present also in medical alignment, and we propose a modality-agnostic framework that closes this gap, ensuring that semantically related representations are more aligned, regardless of their source modality. Our method enhances alignment between radiology images and clinical text, improving cross-modal retrieval and image captioning.

Giordano Cicchetti, Eleonora Grassucci, Danilo Comminiello

Multimodal learning plays a pivotal role in advancing artificial intelligence systems by incorporating information from multiple modalities to build a more comprehensive representation. Despite its importance, current state-of-the-art models still suffer from severe limitations that prevent the successful development of a fully multimodal model. Such methods may not provide indicators that all the involved modalities are effectively aligned. As a result, some modalities may not be aligned, undermining the effectiveness of the model in downstream tasks where multiple modalities should provide additional information that the model fails to exploit. In this paper, we present TRIANGLE: TRI-modAl Neural Geometric LEarning, the novel proposed similarity measure that is directly computed in the higher-dimensional space spanned by the modality embeddings. TRIANGLE improves the joint alignment of three modalities via a triangle-area similarity, avoiding additional fusion layers or pairwise similarities. When incorporated in contrastive losses replacing cosine similarity, TRIANGLE significantly boosts the performance of multimodal modeling, while yielding interpretable alignment rationales. Extensive evaluation in three-modal tasks such as video-text and audio-text retrieval or audio-video classification, demonstrates that TRIANGLE achieves state-of-the-art results across different datasets improving the performance of cosine-based methods up to 9 points of Recall@1.

Eleonora Grassucci, Giuliano Galadini, Giordano Cicchetti et al.

Video-to-audio (V2A) generation aims to synthesize realistic and semantically aligned audio from silent videos, with potential applications in video editing, Foley sound design, and assistive multimedia. Although the excellent results, existing approaches either require costly joint training on large-scale paired datasets or rely on pairwise similarities that may fail to capture global multimodal coherence. In this work, we propose a novel training-free multimodal guidance mechanism for V2A diffusion that leverages the volume spanned by the modality embeddings to enforce unified alignment across video, audio, and text. The proposed multimodal diffusion guidance (MDG) provides a lightweight, plug-and-play control signal that can be applied on top of any pretrained audio diffusion model without retraining. Experiments on VGGSound and AudioCaps demonstrate that our MDG consistently improves perceptual quality and multimodal alignment compared to baselines, proving the effectiveness of a joint multimodal guidance for V2A.

Eleonora Grassucci, Giordano Cicchetti, Aurelio Uncini et al.

Multimodal representation learning produces high-dimensional embeddings that align diverse modalities in a shared latent space. While this enables strong generalization, it also introduces scalability challenges, both in terms of storage and downstream processing. A key open problem is how to achieve semantic compression, reducing the memory footprint of multimodal embeddings while preserving their ability to represent shared semantic content across modalities. In this paper, we prove a strong connection between reducing the modality gap, which is the residual separation of embeddings from different modalities, and the feasibility of post-training semantic compression. When the gap is sufficiently reduced, embeddings from different modalities but expressing the same semantics share a common portion of the space. Therefore, their centroid is a faithful representation of such a semantic concept. This enables replacing multiple embeddings with a single centroid, yielding significant memory savings. We propose a novel approach for semantic compression grounded on the latter intuition, operating directly on pretrained encoders. We demonstrate its effectiveness across diverse large-scale multimodal downstream tasks. Our results highlight that modality alignment is a key enabler for semantic compression, showing that the proposed approach achieves significant compression without sacrificing performance.