We introduce MAIA (Multimodal AI Assessment), a native-Italian benchmark designed for fine-grained investigation of the reasoning abilities of visual language models on videos. MAIA differs from other available video benchmarks for its design, its reasoning categories, the metric it uses and the language and culture of the videos. It evaluates Vision Language Models (VLMs) on two aligned tasks: a visual statement verification task, and an open-ended visual question-answering task, both on the same set of video-related questions. It considers twelve reasoning categories that aim to disentangle language and vision relations by highlight when one of two alone encodes sufficient information to solve the tasks, when they are both needed and when the full richness of the short video is essential instead of just a part of it. Thanks to its carefully taught design, it evaluates VLMs' consistency and visually grounded natural language comprehension and generation simultaneously through an aggregated metric. Last but not least, the video collection has been carefully selected to reflect the Italian culture and the language data are produced by native-speakers.

Cross-domain fake news detection aims to mitigate domain shift and improve detection performance by transferring knowledge across domains. Existing approaches transfer knowledge based on news content and user engagements from a source domain to a target domain. However, these approaches face two main limitations, hindering effective knowledge transfer and optimal fake news detection performance. Firstly, from a micro perspective, they neglect the negative impact of veracity-irrelevant features in news content when transferring domain-shared features across domains. Secondly, from a macro perspective, existing approaches ignore the relationship between user engagement and news content, which reveals shared behaviors of common users across domains and can facilitate more effective knowledge transfer. To address these limitations, we propose a novel macro- and micro- hierarchical transfer learning framework (MMHT) for cross-domain fake news detection. Firstly, we propose a micro-hierarchical disentangling module to disentangle veracity-relevant and veracity-irrelevant features from news content in the source domain for improving fake news detection performance in the target domain. Secondly, we propose a macro-hierarchical transfer learning module to generate engagement features based on common users' shared behaviors in different domains for improving effectiveness of knowledge transfer. Extensive experiments on real-world datasets demonstrate that our framework significantly outperforms the state-of-the-art baselines.

Detecting musical versions (different renditions of the same piece) is a challenging task with important applications. Because of the ground truth nature, existing approaches match musical versions at the track level (e.g., whole song). However, most applications require to match them at the segment level (e.g., 20s chunks). In addition, existing approaches resort to classification and triplet losses, disregarding more recent losses that could bring meaningful improvements. In this paper, we propose a method to learn from weakly annotated segments, together with a contrastive loss variant that outperforms well-studied alternatives. The former is based on pairwise segment distance reductions, while the latter modifies an existing loss following decoupling, hyper-parameter, and geometric considerations. With these two elements, we do not only achieve state-of-the-art results in the standard track-level evaluation, but we also obtain a breakthrough performance in a segment-level evaluation. We believe that, due to the generality of the challenges addressed here, the proposed methods may find utility in domains beyond audio or musical version matching.

Legendary comedian and actor Kevin Nealon performed on "Saturday Night Live" for almost a decade, acting in some of the series' most iconic sketches. After 40 years in the business, he recently spoke with Fox News Digital about the current state of stand-up comedy and where he feels the industry is headed. Though the medium has evolved into something bigger than ever before, Nealon described the attention spans of modern comedy audiences as much shorter -- something that those involved in the business of humor have had to cater to. "When I started comedy, it was totally different. And it was a totally different time and generation. And it was not as much short attention span. Like, I look back at some of the sketches on'SNL,' and they're a lot longer than they are now because of the short attention span, and a lot of people don't watch'SNL' at that time. They watch it on YouTube, snippets of it," said the comedian, pointing to social media as something that's gotten hundreds of millions of people accustomed to consuming content in short clips and blurbs.

Sir Jony Ive, the innovative designer of Apple's iMac, iPhone and Apple Watch, and a close friend and collaborator of the late Steve Jobs, says he still often asks himself: "I wonder what Steve would do?" Ive told BBC Radio 4's Desert Island Discs on Sunday that he does so despite the fact that Jobs had specifically told him not to before his death in 2011, aged 56. "He used to say I really don't want you to be thinking'Well, what would Steve do?'," Ives said. The designer, who was born in Chingford, Essex, and moved to San Francisco to work at Apple in 1992, worked alongside the company's co-founder and CEO five years later, when Jobs was called back in to help the struggling company after a period working elsewhere. Jobs' return marked an immediate improvement for Ive, he recalled. "It was remarkable that, despite the limitations of my ability to communicate, Steve understood what I thought and how I felt," Ive said.

Target speaker extraction focuses on isolating a specific speaker's voice from an audio mixture containing multiple speakers. To provide information about the target speaker's identity, prior works have utilized clean audio examples as conditioning inputs. However, such clean audio examples are not always readily available (e.g. It is impractical to obtain a clean audio example of a stranger's voice at a cocktail party without stepping away from the noisy environment). Limited prior research has explored extracting the target speaker's characteristics from noisy audio examples, which may include overlapping speech from disturbing speakers. In this work, we focus on target speaker extraction when multiple speakers are present during the enrollment stage, through leveraging differences between audio segments where the target speakers are speaking (Positive Enrollments) and segments where they are not (Negative Enrollments). Experiments show the effectiveness of our model architecture and the dedicated pretraining method for the proposed task. Our method achieves state-of-the-art performance in the proposed application settings and demonstrates strong generalizability across challenging and realistic scenarios.

Fine-tuning pre-trained transformers is a powerful technique for enhancing the performance of base models on specific tasks. From early applications in models like BERT to fine-tuning Large Language Models (LLMs), this approach has been instrumental in adapting general-purpose architectures for specialized downstream tasks. Understanding the fine-tuning process is crucial for uncovering how transformers adapt to specific objectives, retain general representations, and acquire task-specific features. This paper explores the underlying mechanisms of fine-tuning, specifically in the BERT transformer, by analyzing activation similarity, training Sparse AutoEncoders (SAEs), and visualizing token-level activations across different layers. Based on experiments conducted across multiple datasets and BERT layers, we observe a steady progression in how features adapt to the task at hand: early layers primarily retain general representations, middle layers act as a transition between general and task-specific features, and later layers fully specialize in task adaptation. These findings provide key insights into the inner workings of fine-tuning and its impact on representation learning within transformer architectures.

Generative AI release decisions determine whether system components are made available, but release does not address many other elements that change how users and stakeholders are able to engage with a system. Beyond release, access to system components informs potential risks and benefits. Access refers to practical needs, infrastructurally, technically, and societally, in order to use available components in some way. We deconstruct access along three axes: resourcing, technical usability, and utility. Within each category, a set of variables per system component clarify tradeoffs. For example, resourcing requires access to computing infrastructure to serve model weights. We also compare the accessibility of four high performance language models, two open-weight and two closed-weight, showing similar considerations for all based instead on access variables. Access variables set the foundation for being able to scale or increase access to users; we examine the scale of access and how scale affects ability to manage and intervene on risks. This framework better encompasses the landscape and risk-benefit tradeoffs of system releases to inform system release decisions, research, and policy.

The proliferation of multimodal content on social media presents significant challenges in understanding and moderating complex, context-dependent issues such as misinformation, hate speech, and propaganda. While efforts have been made to develop resources and propose new methods for automatic detection, limited attention has been given to label detection and the generation of explanation-based rationales for predicted labels. To address this challenge, we introduce MemeIntel, an explanation-enhanced dataset for propaganda memes in Arabic and hateful memes in English, making it the first large-scale resource for these tasks. To solve these tasks, we propose a multi-stage optimization approach and train Vision-Language Models (VLMs). Our results demonstrate that this approach significantly improves performance over the base model for both \textbf{label detection} and explanation generation, outperforming the current state-of-the-art with an absolute improvement of ~3% on ArMeme and ~7% on Hateful Memes. For reproducibility and future research, we aim to make the MemeIntel dataset and experimental resources publicly available.

Recent advancements in audio tokenization have significantly enhanced the integration of audio capabilities into large language models (LLMs). However, audio understanding and generation are often treated as distinct tasks, hindering the development of truly unified audio-language models. While instruction tuning has demonstrated remarkable success in improving generalization and zero-shot learning across text and vision, its application to audio remains largely unexplored. A major obstacle is the lack of comprehensive datasets that unify audio understanding and generation. To address this, we introduce Audio-FLAN, a large-scale instruction-tuning dataset covering 80 diverse tasks across speech, music, and sound domains, with over 100 million instances. Audio-FLAN lays the foundation for unified audio-language models that can seamlessly handle both understanding (e.g., transcription, comprehension) and generation (e.g., speech, music, sound) tasks across a wide range of audio domains in a zero-shot manner. The Audio-FLAN dataset is available on HuggingFace and GitHub and will be continuously updated.