Large reasoning models (LRMs) have recently achieved significant progress in complex reasoning tasks, aided by reinforcement learning with verifiable rewards. However, LRMs often suffer from overthinking, expending excessive computation on simple problems and reducing efficiency. Existing efficient reasoning methods typically require accurate task assessment to preset token budgets or select reasoning modes, which limits their flexibility and reliability. In this work, we revisit the essence of overthinking and identify that encouraging effective steps while penalizing ineffective ones is key to its solution. To this end, we propose a novel rule-based verifiable stepwise reward mechanism (VSRM), which assigns rewards based on the performance of intermediate states in the reasoning trajectory. This approach is intuitive and naturally fits the step-by-step nature of reasoning tasks. We conduct extensive experiments on standard mathematical reasoning benchmarks, including AIME24 and AIME25, by integrating VSRM with PPO and Reinforce++. Results show that our method achieves substantial output length reduction while maintaining original reasoning performance, striking an optimal balance between efficiency and accuracy. Further analysis of overthinking frequency and pass@k score before and after training demonstrates that our approach indeed effectively suppresses ineffective steps and encourages effective reasoning, fundamentally alleviating the overthinking problem.

Metal artifacts frequently emerge in reconstructed computerized tomography (CT) images of cultural relics, primarily due to the presence of metal elements within these cultural relics during the scanning process. It degrades the quality of CT images, posing a challenge to cultural relic restoration efforts. Traditional metal artifact reduction (MAR) methods include projection-domain interpolation and iterative reconstruction. The former offers high computational efficiency but may introduce secondary artifacts, while the latter achieves high accuracy at the cost of increased computational complexity. Recent studies have employed CNN-based image segmentation models to automatically separate metal artifacts from anatomical structures in medical CT images. These models demonstrate superior accuracy compared to conventional approaches. However, compared to medical CT images, the CT images of cultural relics are more complex and blurred at the edges due to their age and the diversity of their constituent materials. This makes artifact removal in cultural relics a challenging task. To address this problem, we first constructed a dataset of CT images of Chinese cultural relics and annotated the cultural relic regions in each image. Secondly, we classified the images into five classes based on the shape of the artifact, then proposed the Annotation and Structural Variability Estimator (ASVE) to assess the complexity of each class for division of test set and training set. Finally, we proposed a CT image artifact removal framework for cultural relics, utilizing a domain-specific pre-trained encoder and a designed attention-based scSEConv Block to enhance the artifact awareness and boundary recovery capabilities of the Unet. The experimental results showed that our model achieved an IOU of 0.8445 and a recall of 0.9179 on the test set, outperforming other compared segmentation models.

Metagenomic data, comprising mixed multi-species genomes, are prevalent in diverse environments like oceans and soils, significantly impacting human health and ecological functions. However, current research relies on K-mer, which limits the capture of structurally and functionally relevant gene contexts. Moreover, these approaches struggle with encoding biologically meaningful genes and fail to address the one-to-many and many-to-one relationships inherent in metagenomic data. To overcome these challenges, we introduce FGeneBERT, a novel metagenomic pre-trained model that employs a protein-based gene representation as a context-aware and structure-relevant tokenizer. FGeneBERT incorporates masked gene modeling to enhance the understanding of inter-gene contextual relationships and triplet enhanced metagenomic contrastive learning to elucidate gene sequence-function relationships. Pre-trained on over 100 million metagenomic sequences, FGeneBERT demonstrates superior performance on metagenomic datasets at four levels, spanning gene, functional, bacterial, and environmental levels and ranging from 1 to 213 k input sequences. Case studies of ATP synthase and gene operons highlight FGeneBERT's capability for functional recognition and its biological relevance in metagenomic research.

Computation offloading is one of the key issues in mobile edge computing (MEC) that alleviates the tension between user equipment's limited capabilities and mobile application's high requirements. To achieve model-free computation offloading when reliable MEC dynamics are unavailable, deep reinforcement learning (DRL) has become a popular methodology. However, most existing DRL-based offloading approaches are developed for a single MEC environment, with invariant system bandwidth, edge capability, task types, etc., while realistic MEC scenarios tend to be of high diversity. Unfortunately, in multi-MEC environments, DRL-based offloading faces at least two challenges, learning inefficiency and interference of offloading experiences. To address the challenges, we propose a DRL-based Multi-environmental Module-compositional Modelfree computation OFFloading (M3OFF) framework. M3OFF generates offloading policies using module composition instead of a single DRL network so that learning efficiency could be improved by reusing the same modules and learning interference could be reduced by composing different modules. Furthermore, we design multiple module composition-specific training methods for M3OFF, including alternate modules-and-composer updates to improve training stability, loss-regularization to avoid module degeneration, and module-dropout to mitigate overfitting. Extensive experimental results on both simulation and testbed demonstrate that M3OFF outperforms the performances of most state-of-the-arts in multi-MEC and reaches close to single-MEC.

Abstract Summary: NMPP package is a bundle of user-customized tools based on established algorithms and methods to process self-designed NimbleGen microarray data. It features a command-line-based integrative processing procedure that comprises five major functional components, namely the raw microarray data parsing and integrating module, the array spatial effect smoothing and visualization module, the probe-level multi-array normalization module, the gene expression intensity summarization module and the gene expression status inference module. Availability: Contact: [email protected] Supplementary information: The supplementary materials, figures, testing dataset, software instructions and a typical workflow of using NMPP package are accessible through the above homepage.