Learning with softmax cross-entropy on one-hot labels often leads to overconfidence on the correct class. While label smoothing regulates this overconfidence by redistributing some confidence from the correct class to other incorrect classes, it compromises the representation in the logits about the similarity between samples of different classes and may hurt calibration if higher confidence is required for high accuracy. To overcome these limitations, we propose a Virtual Smoothing (VS) label that redistributes certain confidence from the correct class to additional VS classes to regularize overconfidence. In VS labels, the VS class nodes act as adversaries to the original class nodes, enforcing regularization by clustering samples across all classes. The zero confidence assigned to each incorrect class also allows the incorrect logits to be different from each other without erasing information about sample similarities. The prediction probability can still approach 1 when applying softmax to the logits of the original real classes, which avoids harming but consistently improves calibration. Experiments show that VS labels consistently improve accuracy and calibration while providing better logits for improved knowledge distillation. Additionally, VS labels exhibit effectiveness in improving adversarial training, robust distillation, and out-of-distribution detection.

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.

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.

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.

Big data processing is now the primary mission in remote sensing processing, fortunately, cloud computing provides a feasible approach to perform it efficiently. But the work of resource provisioning, scheduling, and scaling is still inevitable in most cloud computing solutions, it poses a considerable challenge to data analyst. The emerging serverless architecture presents a new paradigm to provide a cloud service, the user only needs to upload function codes and leaves all the other server management jobs to the service provider. It reveals a new possibility of remote sensing processing. This paper presents FaasRS, a framework to process remote sensing images upon serverless platform. FaasRS is built on AWS Lambda, it exposes only simple APIs to operate images, and builds DAG for user’s algorithm. FaasRS splits task by splitting the image into small tiles based on geospatial region, and uses each Lambda worker to perform the computation for one tile. To reduce the redundant operations, we also make optimizations based on the algorithm DAG. FaasRS shows favorable performance and scalability in our evaluation. In the comparison with Spark and Ray, FaasRS shows a significant performance improvement in different type of RS processing jobs.