The reactive oxygen species (O*) released from the Nickel-rich layered oxide cathodes (LiNi_xCo_yMn_1−x−yO₂, NCM) are responsible for triggering thermal runaway (TR) in lithium-ion batteries (LIBs). Specifically, the charge compensation from transition metal (TM) 3d to oxygen (O) 2p in NCM plays a pivotal role in O* release. Here, inspired by the strong chelating effect of sodium phytate (PN) on TM, this study utilizes PN as a cathode additive to interact with nickel, weaken the charge compensation of TM 3d to O 2p on the surface of LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811) and enhance the battery safety. It is shown that the chelation successfully stabilizes lattice oxygen and inhibits O* release, preventing harmful phase transitions in NCM811 and attenuating heat generation from O* related crosstalk reactions. Consequently, the TR trigger temperature (T_tr) of NCM811 pouch cell with PN increases from 125.9 to 184.8 °C, while the maximum temperature (T_max) decreases from 543.7 to 319.7 °C. Moreover, the PN-modified layer allows NCM811 to be cycled stably for over 700 cycles at 4.6 V. This strategy provides a facile method for stabilizing lattice oxygen in NCM, inhibiting O*-triggered TR, and enhancing high-voltage performance.

Human-robot interaction (HRI) depends on advanced sensing technologies to ensure both safety and efficiency. However, most current robotic sensors offer limited functionality. This study presents a fully soft robotic sensor with four integrated sensing capabilities: spatial proximity sensing, non-contact thermal sensing, contact-based thermal sensing, and mechanical force sensing. This multipurpose sensor enables precise detection of thermal and mechanical stimuli in both contact and non-contact manners. When integrated with a soft gripper and robotic arm, the robotic sensor demonstrated robust performance across a range of HRI scenarios. This technique could advance robotic perception and adaptability in complex environments.

This review covers recent advances in artificial intelligence (AI) for aquatic species identification and conservation, systematically summarizing the current applications and challenges of machine learning algorithms in image analysis, acoustic identification, and ecological threat detection. It highlights that AI models such as convolutional neural network (CNN) and recurrent neural network (RNN) have demonstrated significantly higher accuracy and scalability in species classification, habitat monitoring, and environmental threat detection compared to traditional ecological survey methods. Additionally, through bibliometric analysis, the study identifies global research trends and interdisciplinary collaboration patterns, emphasizing the importance of strengthening cross-disciplinary cooperation, establishing standardized protocols, and developing open data platforms to support sustainable aquatic ecosystem conservation.

This study proposes an automatic liver segmentation method for computed tomography (CT) images based on an improved ResUNet, which integrates the advantages of UNet and ResNet architectures to achieve high-precision segmentation on 128×128pixel images. The model removes batch normalization layers and incorporates residual blocks, combined with data augmentation techniques, ultimately achieving excellent metrics such as a Dice coefficient of 93.08% and accuracy of 98.57%. Surpassing traditional methods, this approach provides efficient and reliable technical support for liver disease diagnosis.