Video-based Sign Language Recognition (SLR) is a fundamental accessibility technology, yet real-world sign vocabularies follow an extreme long-tailed distribution, a problem sharpened for low-resource languages such as Vietnamese Sign Language (VSL), where classical imbalance tools (re-weighting, focal loss, GAN-based augmentation) either over-fit the head or leave the tail under-determined. We ask whether the bottleneck in this regime is representation quality or the under-determination of per-class decision boundaries, and whether a generative prior over class prototypes—rather than over raw video—can supply the missing constraint. We propose LT-SignDiff, which repurposes a conditional diffusion model as a generative prior in feature space: it synthesizes a discriminative prototype for each rare class directly in a skeleton encoder's embedding space, conditioned on the k nearest frequent-class prototypes, and injects it into a hybrid cosine classifier. The distinct contribution is that synthesis is performed in prototype space and conditioned on a local semantic neighbourhood—unlike prior diffusion methods that generate pixel-space samples or a single global feature distribution—which can be interpreted as a shrinkage estimator that stabilizes one-shot classifier rows. On a 3,782-class VSL benchmark whose original distribution is extreme (1.60 recordings per class on average, 64% of classes with a single example), LT-SignDiff attains 62.5% Top-1 / 81.4% Top-5 / 58.2% Macro-F1, improving over the strongest non-generative baseline by +13.3 Top-1 / +12.4 Macro-F1 and over the best prior VSL baseline by +32.7 Top-1. Ablations attribute the majority of the tail-class gain to the diffusion prototype synthesis, indicating that for low-resource SLR the most effective place to be generative is the prototype space, not the pixel space.
Active Speaker Detection (ASD) aims to identify the speaking individual within a visual scene across consecutive frames. This task has gained significant prominence due to its role in applications like Human-Robot Interaction (HRI). While existing multimodal approaches primarily focus on innovating video encoders, leveraging architectures such as Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Graph Neural Networks (GNN), audio encoders have remained almost unaltered, often relying on Mel-frequency cepstral coefficients (MFCCs) processed through standard CNNs. In this paper, we explore the use of alternative audio encoding architectures and fusion mechanisms within Light-ASD framework. We conduct an extensive comparative analysis under in-database and cross-database settings using the AVA and WASD datasets. Furthermore, we investigate the impact of multi-dataset training by merging the training sets of both. Our results demonstrate that under in-database conditions, the MFCCs model achieves the best performance for AVA (93.54% mAP), while Wav2Vec yields the best results for WASD (94.28% mAP). In cross-database evaluations, X-Vectors exhibits the best generalization capabilities when trained on AVA (87.35% mAP, with a -2.76% relative performance drop relative to the in-database setting), whereas Wav2Vec generalizes more effectively when trained on WASD (83.62% mAP, with a -8.03% relative performance drop). Finally, we show that dataset fusion significantly enhances model robustness, delivering performance metrics that are comparable to their in-database setting, surpassing them in certain instances.
Recent advances in large language models (LLMs) have enabled impressive performance across reasoning tasks, yet their capacity for autonomous moral judgment remains underexplored. Prior research has not systematically benchmarked LLMs against expert standards of postconventional moral reasoning. This study aims to evaluate the moral reasoning capabilities of twelve state-of-the-art LLMs by comparing their responses to complex socio-moral dilemmas with those of a moral psychology expert, using Kohlberg’s framework as a referential standard. Models were prompted with four dilemmas (passive euthanasia, incurable diseases decisions, population control, and capital punishment) across three zero-shot tasks: moral recognition, option selection, and justification. Responses were coded using Kohlberg’s levels of moral development and statistically compared to expert judgments. Quantitative metrics included weighted kappa for concordance and logistic regression for explanatory depth. Qualitative analysis examined the structure and coherence of LLM justifications. LLMs predominantly exhibited conventional reasoning (47.7%), diverging sharply from the expert’s consistently postconventional judgment. Concordance was weak to negative, with kappa coefficients ranging from –0.333 to 0.286, indicating systemic misalignment. Binary logistic regression analysis showed that model classification significantly predicted the capacity for moral articulation; the odds ratio was 7.00 (p = 0.005), accounting for 23.4% of the variance according to Nagelkerke’s R² = 0.234. The Kruskal–Wallis test yielded H = 20.34 with p < .05, confirming significant differences in moral reasoning levels between the models and the expert. Finally, a qualitative analysis revealed that although 84.3% of justifications aligned with the selected options, they often lacked principled depth, replicating normative patterns and rejecting postconventional stances through lower-order reasoning. In conclusion, contemporary LLMs reproduce social conventions rather than engaging in autonomous ethical deliberation. The observed limitations have substantive implications for the ethical design and deployment of AI systems in morally sensitive domains. To address this, the study proposes six complementary strategies for ethical alignment, aiming to foster principled moral reasoning in AI systems: staged fine-tuning, symbolic ontology integration, expert supervision, reinforcement learning with philosophical rewards, adversarial training, and continual calibration.
Flood disasters pose severe threats to human lives and infrastructure worldwide. Accurate and timely flood detection using satellite imagery is essential for emergency response and disaster management. This paper introduces MUST-Former, a family of five deep learning models designed for robust global flood mapping using multi-modal satellite data. The proposed architectures leverage Sentinel-1 radar and Sentinel-2 optical imagery from the Sen1Floods11 dataset, which covers 11 flood-affected regions across 6 continents. Three distinct fusion strategies are presented: Projection-based fusion, Cross-Attention fusion, and the flagship Principal Component Analysis-inspired fusion. The models address critical challenges in flood detection, including severe class imbalance, geographic bias, and sensor-specific limitations. The MUST-Former-PCA variant achieves a mean Intersection over Union of 90.35% on the test set and 89.32% on the unseen Bolivia region, demonstrating superior generalization. Compared to existing baselines, MUST-Former-SAR shows a 160% improvement in mean IoU over the original Sen1Floods11 baseline, while MUST-Former-Optical achieves competitive performance against foundation models using 3.6 times fewer parameters and superior geographic generalization on the unseen Bolivia split. The fusion models successfully mitigate single-modality failures across diverse geographic regions. Statistical reproducibility is confirmed across three independent random seeds. These results prove the effectiveness of multi-modal approaches for operational flood mapping systems.
Decentralized Autonomous Organizations (DAOs) offer a novel approach to collectively governing projects through a democratic mechanism facilitated by blockchain. DAOs allow members to put forward and vote on proposals, thereby shaping the organization’s future. However, low voter turnout is common in DAO decision-making, particularly in large and active DAOs, where the high volume of proposals makes it unrealistic to expect members to track all proposals. Abstentionism threatens the voting system's effectiveness and the legitimacy of the results. We consider that recommender systems can help boost voter engagement. This article details the design of a recommendation approach tailored to aid DAO members in identifying proposals of interest, alongside its implementation and evaluation. The design accommodates the domain constraints that render off-the-shelf recommendation approaches inadequate, namely that proposals are short-lived and can only be recommended while available for voting. To the best of our knowledge, this is the first study to examine recommendation in DAO governance. To carry out our research, we have compiled a dataset, made publicly available, covering 12 of the most active DAOs. We compare a baseline, specifically designed to accommodate DAO-specific constraints, against a range of recommendation techniques. The findings confirm that personalized recommenders can often anticipate voting preferences, significantly outperforming the baseline. In turn, given the limitations of offline evaluation, an online evaluation using A/B testing would also be needed to fully assess their impact on participation. We also discuss how their implementation must carefully incorporate fairness and transparency to ensure community trust and adoption. We believe that proposal recommender systems in DAOs not only could drive engagement improvements similar to those observed in other online collaborative projects, but can also provide insights for collective governance settings beyond blockchain, such as cooperative organizations or participatory budgeting platforms.
