The authors claim that there is no suitable metric to evaluate the quality of the generated audio, which is plausible, so they listened to the audio and evaluated on their own. The only shortcoming here is that no systematic and blind listening test has been conducted yet. The authors themselves might be biased and thus, the capabilities of the proposed approach cannot be considered as fully proven from a scientific perspective. However, a link to the audio is provided so that the readers can convince themselves from the proposed method. Minor comments: -"nats per timestep": should be defined -p. 3, l.

How are we able to learn about complex current events just from short snippets of video? While natural language enables straightforward ways to represent under-specified, partially observable events, visual data does not facilitate analogous methods and, consequently, introduces unique challenges in event understanding. With the growing prevalence of vision-capable AI agents, these systems must be able to model events from collections of unstructured video data. To tackle robust event modeling in multimodal settings, we introduce a multimodal formulation for partially-defined events and cast the extraction of these events as a three-stage span retrieval task. We propose a corresponding benchmark for this task, MultiVENT-G, that consists of 14.5 hours of densely annotated current event videos and 1,168 text documents, containing 22.8K labeled event-centric entities. We propose a collection of LLM-driven approaches to the task of multimodal event analysis, and evaluate them on MultiVENT-G. Results illustrate the challenges that abstract event understanding poses and demonstrates promise in event-centric video-language systems.

Large text-to-image diffusion models Saharia et al. (2022); Pernias et al. (2024); Podell et al. (2024); Ramesh et al. (2022) have demonstrated significant capabilities in generating photorealistic images based on given textual prompts, facilitating both the creation and editing of real images. Current research Cao et al. (2023); Brack et al. (2024); Ju et al. (2024); Parmar et al. (2023); Wu & la Torre (2022); Xu et al. (2024) highlights three main challenges in image editing: controllability, background preservation, and efficiency. Specifically, the edited parts must align with the target prompt's concepts, while unedited regions should remain unchanged. Additionally, the editing process must be sufficiently efficient to support interactive tasks. There are two mainstream categories of image editing approaches, namely inversion-based and inversion-free methods, as illustrated in Figure 1. Inversion-based approaches Song et al. (2021a); Mokady et al. (2023); Wu & la Torre (2022); Huberman-Spiegelglas et al. (2024) progressively add noise to a clean image and then remove the noise conditioned on a given target prompt, utilizing large text-to-image diffusion models (i.e. Stable Diffusion Rombach et al. (2022)), to obtain the edited image. However, directly inverting the diffusion sampling process (e.g., DDIM Song et al. (2021a)) for reconstruction introduces bias from the initial image due to errors accumulated by an unconditional score term, as discussed in classifier-free guidance (CFG) Ho & Salimans (2022) and proven in App.

We present a novel approach to classify causal micro-narratives from text. These narratives are sentence-level explanations of the cause(s) and/or effect(s) of a target subject. The approach requires only a subject-specific ontology of causes and effects, and we demonstrate it with an application to inflation narratives. Using a human-annotated dataset spanning historical and contemporary US news articles for training, we evaluate several large language models (LLMs) on this multi-label classification task. The best-performing model--a fine-tuned Llama 3.1 8B--achieves F1 scores of 0.87 on narrative detection and 0.71 on narrative classification. Comprehensive error analysis reveals challenges arising from linguistic ambiguity and highlights how model errors often mirror human annotator disagreements. This research establishes a framework for extracting causal micro-narratives from real-world data, with wide-ranging applications to social science research.

The Greenland Ice Sheet (GrIS) has emerged as a significant contributor to global sea level rise, primarily due to increased meltwater runoff. Supraglacial lakes, which form on the ice sheet surface during the summer months, can impact ice sheet dynamics and mass loss; thus, better understanding these lakes' seasonal evolution and dynamics is an important task. This study presents a computationally efficient time series classification approach that uses Gaussian Mixture Models (GMMs) of the Reconstructed Phase Spaces (RPSs) to identify supraglacial lakes based on their seasonal evolution: 1) those that refreeze at the end of the melt season, 2) those that drain during the melt season, and 3) those that become buried, remaining liquid insulated a few meters beneath the surface. Our approach uses time series data from the Sentinel-1 and Sentinel-2 satellites, which utilize microwave and visible radiation, respectively. Evaluated on a GrIS-wide dataset, the RPS-GMM model, trained on a single representative sample per class, achieves 85.46% accuracy with Sentinel-1 data alone and 89.70% with combined Sentinel-1 and Sentinel-2 data. This performance significantly surpasses existing machine learning and deep learning models which require a large training data. The results demonstrate the robustness of the RPS-GMM model in capturing the complex temporal dynamics of supraglacial lakes with minimal training data.

We propose a novel approach for humming transcription that combines a CNN-based architecture with a dynamic programming-based post-processing algorithm, utilizing the recently introduced HumTrans dataset. We identify and address inherent problems with the offset and onset ground truth provided by the dataset, offering heuristics to improve these annotations, resulting in a dataset with precise annotations that will aid future research. Additionally, we compare the transcription accuracy of our method against several others, demonstrating state-of-the-art (SOTA) results. All our code and corrected dataset is available at https://github.com/shubham-gupta-30/humming_transcription

In this article, we propose and apply a method to compare adaptations of the same story across different media. We tackle this task by modelling such adaptations through character networks. We compare them by leveraging two concepts at the core of storytelling: the characters involved, and the dynamics of the story. We propose several methods to match characters between media and compare their position in the networks; and perform narrative matching, i.e. match the sequences of narrative units that constitute the plots. We apply these methods to the novel series \textit{A Song of Ice and Fire}, by G.R.R. Martin, and its comics and TV show adaptations. Our results show that interactions between characters are not sufficient to properly match individual characters between adaptations, but that using some additional information such as character affiliation or gender significantly improves the performance. On the contrary, character interactions convey enough information to perform narrative matching, and allow us to detect the divergence between the original novels and its TV show adaptation.

Forensic analysts often use social media imagery and texts to understand important events. A primary challenge is the initial sifting of irrelevant posts. This work introduces an interactive process for training an event-centric, learning-based multimodal classification model that automates sanitization. We propose a method based on Bayesian Graph Neural Networks (BGNNs) and evaluate active learning and pseudo-labeling formulations to reduce the number of posts the analyst must manually annotate. Our results indicate that BGNNs are useful for social-media data sifting for forensics investigations of events of interest, the value of active learning and pseudo-labeling varies based on the setting, and incorporating unlabelled data from other events improves performance.

Despite advances in diffusion-based text-to-music (TTM) methods, efficient, high-quality generation remains a challenge. We introduce Presto!, an approach to inference acceleration for score-based diffusion transformers via reducing both sampling steps and cost per step. To reduce steps, we develop a new score-based distribution matching distillation (DMD) method for the EDM-family of diffus ion models, the first GAN-based distillation method for TTM. To reduce the cost per step, we develop a simple, but powerful improvement to a recent layer distillation method that improves learning via better preserving hidden state variance. Finally, we combine our step and layer distillation methods together for a dual-faceted approach. We evaluate our step and layer distillation methods independently and show each yield best-in-class performance. Our combined distillation method can generate high-quality outputs with improved diversity, accelerating our base model by 10-18x (230/435ms latency for 32 second mono/stereo 44.1kHz, 15x faster than comparable SOTA) -- the fastest high-quality TTM to our knowledge. We have seen a renaissance of audio-domain generative media (Chen et al., 2024; Agostinelli et al., 2023; Liu et al., 2023; Copet et al., 2023), with increasing capabilities for both Text-to-Audio (TTA) and Text-to-Music (TTM) generation. This work has been driven in-part by audio-domain diffusion models (Song et al., 2020; Ho et al., 2020; Song et al., 2021), enabling considerably better audio modeling than generative adversarial network (GAN) or variational autoencoder (VAE) methods (Dhariwal & Nichol, 2021). Diffusion models, however, suffer from long inference times due to their iterative denoising process, requiring a substantial number of function evaluations (NFE) during inference (i.e.

Large language models (LLMs) perform very well in several natural language processing tasks but raise explainability challenges. In this paper, we examine the effect of random elements in the training of LLMs on the explainability of their predictions. We do so on a task of opinionated journalistic text classification in French. Using a fine-tuned CamemBERT model and an explanation method based on relevance propagation, we find that training with different random seeds produces models with similar accuracy but variable explanations. We therefore claim that characterizing the explanations' statistical distribution is needed for the explainability of LLMs. We then explore a simpler model based on textual features which offers stable explanations but is less accurate. Hence, this simpler model corresponds to a different tradeoff between accuracy and explainability. We show that it can be improved by inserting features derived from CamemBERT's explanations. We finally discuss new research directions suggested by our results, in particular regarding the origin of the sensitivity observed in the training randomness.