Build failures are a major obstacle in RISC-V software migration, often involving complex interactions across logs, configurations, and environments. Traditional diagnostic tools struggle with the unstructured, multi-phase nature of build logs and lack semantic reasoning.We propose a two-stage framework for automated root cause analysis. RV-LAD compresses logs using template-based filtering and applies phase-aware anomaly detection via few-shot LLM prompting. MCTS-RCA integrates a domain-specific knowledge base with Monte Carlo Tree Search to perform LLM-guided multi-source reasoning under classification constraints.To support evaluation, we construct a curated dataset of 117 real-world RISC-V build failures, each annotated with logs, spec files, and repair records. Experiments show our approach achieves 75.2% diagnosis accuracy, surpassing previous LLM-based and rule-based methods. It also offers interpretable reasoning traces, enabling practical and transparent diagnosis. This work provides an effective and extensible solution for RCA in emerging software ecosystems like RISC-V, bridging large language models with domain-aware inference.

Large language models (LLMs) are increasingly integrated into software development workflows to accelerate code generation, but often produce insecure and uncontrollable code due to vulnerable training data and unconstrained decoding strategies. This poses severe risks in security-critical systems, where post-generation vulnerability detection and manual remediation incur significant overhead. While constrained decoding offers a practical mitigation strategy, existing methods suffer from degraded trustworthiness, constraint conflicts, and high latency—especially when enforcing multiple concurrent security constraints.We propose SCodeGen, a real-time constrained decoding framework designed to enforce fine-grained security controls during LLM code generation. To improve trustworthiness and controllability, SCodeGen introduces (1) a matching-length-aware logit modulation strategy that enhances trustworthiness and controllability without semantic disruption, and (2) a two-stage low-latency decoding architecture, which compiles constraint phrases into a runtime-enforceable constraint automaton (RCA) with precomputed logit bias vectors for efficient online decoding. Extensive evaluations on CodeGuard+ show that SCodeGen significantly improves secure pass rates under both single and multi-constraint settings, while maintaining latency comparable to unconstrained decoding. This work demonstrates a practical and scalable solution toward trustworthy LLM-assisted software development under security constraints.

Multimodal sarcasm is often used to express strong emotions online through the discrepancy of the literal-figurative scene across multi-modalities. Current researches retrofit transform-based pretrained language models to integrate text and image to detect sarcasm. However, these methods struggle to distinguish subtle semantic and emotional differences between image and text within the same instance. To address this issue, this paper proposes a new context-aware dual attention network that collaboratively performs textual and visual attentions using a shared memory module. This approach enables us to reason about the interconnected portions involving sarcasm in both text and image. Additionally, we use implicit context derived from multimodal commonsense graph to establish a holistic perspective that encompasses semantics and emotions across modalities. Finally, multi-view cross-modal matching technique is employed to effectively identify contradictions. We evaluate our method on the widely used HFM dataset and achieve 1.01% improvements on the F1-score. Extensive experiments demonstrate the effectiveness of the proposed method.

Within the realms of scientific computing, large-scale data processing, and artificial intelligence-powered computation, disparities in performance, which originate from differing code implementations, directly influence the practicality of the code. Although existing works tried to utilize code knowledge to enhance the execution performance of codes generated by large language models, they neglect code evaluation outcomes which directly refer to the code execution details, resulting in inefficient computation. To address this issue, we propose DSCT-Decode, an innovative adaptive decoding strategy for large language models, that employs a data structure named 'Code Token Tree' (CTT), which guides token selection based on code evaluation outcomes. DSCT-Decode assesses generated code across three dimensions---correctness, performance, and similarity---and utilizes a dynamic penalty-based boundary intersection method to compute multi-objective scores, which are then used to adjust the scores of nodes in the CTT during backpropagation. By maintaining a balance between exploration, through token selection probabilities, and exploitation, through multi-objective scoring, DSCT-Decode effectively navigates the code space to swiftly identify high-performance code solutions. To substantiate our framework, we developed a new benchmark, big-DS-1000, which is an extension of DS-1000. This benchmark is the first of its kind to specifically evaluate code generation methods based on execution performance. Comparative evaluations with leading large language models, such as CodeLlama and GPT-4, show that our framework achieves an average performance enhancement of nearly 30%. Furthermore, 30% of the codes exhibited a performance improvement of more than 20%, underscoring the effectiveness and potential of our framework for practical applications.

Modeling spatial and temporal dependencies is essential for achieving accurate air quality prediction. Current air quality prediction models often overlook the underlying causal relationships in the data and primarily focus on statistical correlations. As a result, these models lack sufficient predictive power for medium- and long-term forecasts. This paper introduces causality into air quality prediction and proposes a Causal Spatio-Temporal Convolutional Network (CSTCN). We utilize an attention mechanism to automatically assign attention weights to each air quality monitoring site, enabling the capture of causal relationships between sites in the spatial dimension. Conducting tests on the identified relationships ensures the causality of the data in space. Furthermore, convolution operations are applied to extract the spatio-temporal features of the monitoring stations, while also utilizing causal convolution to ensure the causality of the data over time. The experiments conducted on the Beijing air quality dataset demonstrate that CSTCN exhibits outstanding performance in medium- and long-term predictions.

Deep Neural Network (DNN) classifiers easily yield high confidence for Out-of-Distribution (OOD) examples beyond the training distribution, i.e., In-Distribution (ID), leading to classification errors. Detecting and rejecting various OOD examples is crucial for the reliability of DNNs. More challenging, well-built detections can also suffer from being re-bypassed by adversarial attacks perturbing unseen OOD examples. Some existing works introduce adversarial training on the auxiliary outliers to improve the robustness of OOD detection. However, in this work, we find that applying adversarial training on the auxiliary outliers is insufficient to make the detection robust to strong adaptive attacks. To fix this bug of OOD detection, we propose a semi-supervised adversarial training approach, RobDet, which mines adversarially perturbed ID examples from within the neighborhood of clean ID ones as auxiliary outliers and uses multiple "other" classes to train them together with other auxiliary clean and adversarially perturbed outliers to enhance the robustness of OOD detection without significantly sacrificing the performance on clean OOD examples. Experiments show that RobDet has a significant advantage in detecting malicious OOD examples generated by strong adaptive attacks while maintaining advanced performance in detecting clean OOD examples.

Environment configuration is the basis for software reuse, enabling developers to reuse specific functions.However, the lack of uniform practice in dependency declaration specifications of Python projects can cause problems for developers trying to install third-party libraries.Existing package management tools are often inadequate to help fix these problems.Fixing these errors requires expensive hours and domain knowledge for developers.To help address related problems, some studies focus on well-maintained and popular Python projects about dependency conflict problems caused by PIP's installation rules.However, many projects in the wild are outside of this scope.We carefully investigate 110 issues in 110 projects in the wild.Based on the comprehensive study, we design and implement EasyPip to automatically detect and fix problems in Python dependency declaration files.Dif-

Channel pruning is a commonly used model compression in convolutional neural network. The structured pruning using sparse constraints can automatically learn the importance of parameters during the training process by imposing sparse constraints on parameters. However, existing pruning methods based on sparse constraints cannot process the final convolutional layer of the residual module with complex connections. Due to the existence of residual connection, if the final convolutional layer of the residual module is pruned, the sparse channel of the feature map from residual connection does not correspond to the feature map from module output, which will cause the parameters to be unable to be pruned. This paper studies this problem and proposes an identity association group pruning algorithm, which we call IGP. IGP groups the parameters and channels that generate the corresponding feature maps, uses Group Lasso to sparse the same group of parameters as a whole, and forces the sparseness of the parameters with sparse correlation to be consistent with each other. Experiments show that when IGP compresses ResNet56 60% parameters, the model performance only drops 0.36 %, which is better than the existing pruning method based on sparse constraints. In the case of high compression ratio, IGP can compresses ResNet-50 compressesed with 87% parameters and the performance drops only 0.76%, which is 5.17 % higher than the existing methods.