Cooperative perception enabled by Vehicle-to-Everything (V2X) communication enhances autonomous driving safety by creating a unified environmental representation through shared sensory data. While recent works have advanced multi-agent fusion for improved perception, uncertainty quantification in such cooperative frameworks remains largely unexplored. This paper introduces Hyper-V2X, a hypernetwork-based framework for estimating both epistemic and aleatoric uncertainties in V2X-based perception. Specifically, we propose a partial weight generation scheme and V2X context embedding module that conditions a Bayesian hypernetwork on fused multi-agent features to generate weight distributions for stochastic Bird's-Eye-View (BEV) segmentation. Unlike existing deterministic BEV models, Hyper-V2X enables efficient uncertainty estimation with little computation overhead. Our approach is architecture-agnostic, and can be seamlessly integrating with modern cooperative backbones such as CoBEVT. Experiments on the OPV2V benchmark demonstrate that Hyper-V2X provides accurate, well-calibrated uncertainty estimates and improves overall perception reliability. Our code and benchmark are publicly available under an open-source license: https://github.com/abhishekjagtap1/Hyper-V2X
Human Trajectory Forecasting (HTF) predicts future human movements from past trajectories and environmental context, with applications in Autonomous Driving, Smart Surveillance, and Human-Robot Interaction. While prior work has focused on accuracy, social interaction modeling, and diversity, little attention has been paid to uncertainty modeling, calibration, and forecasts from short observation periods, which are crucial for downstream tasks such as path planning and collision avoidance. We propose DD-MDN, an end-to-end probabilistic HTF model that combines high positional accuracy, calibrated uncertainty, and robustness to short observations. Using a few-shot denoising diffusion backbone and a dual mixture density network, our method learns self-calibrated residence areas and probability-ranked anchor paths, from which diverse trajectory hypotheses are derived, without predefined anchors or endpoints. Experiments on the ETH/UCY, SDD, inD, and IMPTC datasets demonstrate state-of-the-art accuracy, robustness at short observation intervals, and reliable uncertainty modeling. The code is available at: url{https://github.com/kav-institute/ddmdn}.
Abstract This short editorial introduces the special issue in the Journal of Occupational and Organizational Psychology ‘Workplace coaching at the crossroads’. We outline how workplace coaching has grown into a preferred helping relationship to support workers and leaders in navigating increasingly complex and ever‐changing workplaces. Our definition of coaching frames it as a facilitative, dialectical process that enables positive change and outcomes for the coachee. We summarize that meta‐analytic evidence tells us that coaching ‘works’ by documenting positive impacts on a range of outcomes, including performance, learning and wellbeing. Yet, we know far less about how coaching works and the role of coaching contexts. The five contributions in this special issue address the coaching process. They are methodologically diverse and include both longitudinal and observational data. We posit that the meta‐theme for coaching to work is the coachee's active engagement in shaping the coaching process as they work towards proximal (e.g., self‐awareness) and more distal (e.g., long‐term goals) outcomes. We conclude with a call for future research to address the coaching context by considering micro‐level (e.g., stakeholder alignment), meso‐level (e.g., organizational strategy, coaching culture) and macro‐level considerations (e.g., societal culture, technological transformation).
Acute ischaemic stroke due to large-vessel occlusion requires rapid endovascular treatment (EVT). Prehospital stroke scales, though widely used, have not achieved an optimal balance of sensitivity, specificity, and overall discriminative performance, underlining the need for evaluation of machine learning (ML) approaches to improve early identification of EVT candidates. We conducted a retrospective development and internal validation study based on digital real-world data from consecutive EMS-transported patients with suspected stroke (n = 1,435, 94 EVT events [6.6%]) from the Stroke Angel initiative in a rural region in Germany (Rhön-Grabfeld), covering the period from January 2015 to June 2021. Predicted probabilities of six ML models were recalibrated using Platt scaling, and clinical relevance was evaluated using decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) values were computed to quantify individual feature contributions to model predictions. Among 287 patients (24 EVT events) in the evaluation cohort, Random Forest achieved the highest AUROC (0.85 [95% CI, 0.78–0.90]), with sensitivity of 0.92 [0.74–0.99], specificity of 0.66 [0.60–0.72]. The 4I-SS (score > 2) yielded an AUROC of 0.82 [0.72–0.89], sensitivity of 0.87 [0.67–0.96], and specificity of 0.69 [0.64–0.74]. Bootstrapped intervals overlapped across all metrics, suggesting no statistically significant superiority.
