Hybrid Electric Vehicles (HEVs) has drawn considerable attention, aiming as a practical approach to improving fuel economy and lowering vehicle emissions by combining internal combustion engines with electric propulsion systems. Among the various hybrid designs, P2 and P3 configurations stand out for their unique features related to energy transfer, electric driving capability, and overall powertrain performance. This study compares the performance of P2 and P3 HEV architecture using the SimulationX platform under standardized Worldwide Harmonized Light Vehicles Test Cycle (WLTC) conditions. To make a fair comparison focused on hybrid design, both models used the same vehicle specifications, battery parameters, engine and motor performance characteristics, and control inputs. The performance of both architectures was evaluated using key parameters such as fuel mass flow, battery state of charge (SOC), vehicle velocity, and regenerative braking behavior. The results show that both architectures effectively use electric mode during low-speed conditions below 40 km/h, which reduces fuel consumption during urban driving. Under the investigated WLTC conditions and rule-based energy-management strategy, the P2 architecture recorded lower cumulative fuel consumption and a greater electric-drive-mode contribution than the P3 architecture. P3 exhibited a higher electric-braking-mode contribution. These findings demonstrate that motor placement and its interaction with the selected control strategy influence power flow, operating-mode distribution, and cycle-level energy consumption.
The spread of deepfake videos, created by the application of sophisticated artificial intelligence technologies, presents major hazards to public confidence and digital integrity. To analyze spatial features in individual frames, current deepfake detection techniques mostly depend on convolutional neural networks (CNNs). However, they often overlook temporal dependencies crucial for detecting subtle alterations. We introduce GraphNeXt, a unified framework that combines the spatial representation power of ResNeXt with the temporal modeling capabilities of graph neural networks (GNNs). By integrating three GNN variants (GCN, GAT, and GraphSAGE) with ResNeXt, our approach jointly captures spatial and temporal patterns in video sequences. Our models have been extensively tested on DFDC and Google DFD benchmarks, showcasing superior performance compared to traditional CNN-based approaches, ensuring reliable deepfake detection across a wide range of video origins. These findings underscore the significant role of hybrid CNN-GNN frameworks in progressing video forensics and mitigating the risks posed by synthetic media.
World models are now built on substantially different computational substrates. Latent recurrent state-space models such as PlaNet and the Dreamer family compress observations into recurrent states; token-based models such as IRIS quantize observations into a learned codebook and predict autoregressively with a transformer; and joint-embedding predictive architectures such as I-JEPA predict in a learned latent space with no pixel decoder. The interpretability methods applied to these models, including probing, activation patching, sparse autoencoders, and surprise analysis, share a common set of primitives, yet they are re-implemented from scratch for each architecture because existing hook-and-cache tooling assumes a transformer language model with no notion of actions, environment steps, or imagined rollouts. We argue that this fragmentation reflects the tooling rather than the models, and that the shared structure of world models is captured by a small typed interface. We present WorldModelLens, an open-source interpretability substrate organized around a capability-typed adapter: every model implements four required methods (encode, transition, initial state, sample) and declares a set of optional heads (decode, reward, continue, actor, critic) through an explicit capability descriptor, so that reinforcement-learning and self-supervised world models are first-class without either imitating the other. A single hook and cache layer exposes time-indexed activations, imagination rollouts, and intervention replay over this interface, allowing each analysis to be written once.
Polyester fabrics were functionalized with a phosphorus-based flame retardant, 3-hydroxyphenyl phosphinyl-propanoic acid (3-HPP), and a nitrogen-containing azo disperse dye (Coralene Dark Red 2B) using a high-temperature high-pressure (HTHP) dyeing technique. Two treatment strategies were explored: a one-step process incorporating all components simultaneously during HTHP dyeing, and a two-step method involving dyeing followed by 3-HPP application. Aqueous 3-HPP solutions up to 4% (w/v) were employed. The treatments significantly enhanced flame retardancy, as indicated by an increase in the limiting oxygen index (LOI) from 20.8% (untreated) to 30.0% after plasma activation and 4% 3-HPP application. Co-introduction of phosphorus and nitrogen further improved the LOI to 34.4%, and up to 35.7% after five cycles of plasma treatment, highlighting a synergistic flame-retardant effect. This synergy was supported by X-ray photoelectron spectroscopy (XPS) and energy-dispersive X-ray (EDX) analysis. The treated fabrics demonstrated durable flame resistance, maintaining an LOI of 30.9% after 20 laundering cycles. Vertical flammability tests, thermogravimetric analysis (TGA), and cone calorimetry further confirmed the enhanced thermal stability and flame-retardant performance. Importantly, the mechanical integrity of the fabrics was preserved throughout both plasma and chemical treatments.
Jupyter Notebooks have become the de-facto choice for data scientists and machine learning (ML) engineers for prototyping and experimenting with ML pipelines, offering a rich interactive interface for code, data, and visualization in one place. However, notebooks provide limited support for testing. As a result, during continuous development, many silent (non-crashing) regressions often go unnoticed, making notebooks unreliable and results hard to reproduce. We introduce NBTestGen, the first automated assertion generation approach for ML notebooks. NBTestGen generates regression-based assertions that check properties of data processing, model building, and model evaluation steps in a typical ML notebook. To support integration of such assertions in notebooks, we introduce the first regression testing framework (called NBTest) that can be used as a Jupyter plugin and allows developers to write cell-scoped assertions in notebooks. Such assertions are 1) cell-scoped: they are linked to specific notebook cells and execute only after those cells are executed, 2) non-intrusive: they do not block notebook execution (in a Jupyter session), so that development can continue when they fail, and 3) they integrate with pytest and CI pipelines, allowing developers to easily do regression testing of their notebooks. We evaluate NBTestGen on a corpus of 585 notebooks from the popular Kaggle platform. NBTestGen generates a total of 21542 assertions (36.82 on average per notebook). The generated assertions kill 72.21
