Chain-of-Thought (CoT) prompting has recently shown significant promise across various NLP and computer vision tasks by explicitly generating intermediate reasoning steps. However, we find that reinforcement learning (RL)-based fine-tuned CoT reasoning can paradoxically degrade performance in Visual Grounding tasks, particularly as CoT outputs become lengthy or complex. Additionally, our analysis reveals that increased dataset size does not always enhance performance due to varying data complexities. Motivated by these findings, we propose Curriculum-based Relative Policy Optimization (CuRPO), a novel training strategy that leverages CoT length and generalized Intersection over Union (gIoU) rewards as complexity indicators to progressively structure training data from simpler to more challenging examples. Extensive experiments on RefCOCO, RefCOCO+, RefCOCOg, and LISA datasets demonstrate the effectiveness of our approach. CuRPO consistently outperforms existing methods, including Visual-RFT, reaching a peak improvement of up to 15.49 mAP on RefCOCO. Moreover, CuRPO exhibits exceptional efficiency and robustness, delivering strong localization performance even in few-shot learning scenarios, particularly benefiting tasks characterized by ambiguous and intricate textual descriptions.

Automated Program Repair (APR) aims to automate the patch generation for buggy code and is vital in software devel-opment and maintenance. While large language models (LLMs) excel in various tasks, our empirical study shows they still face challenges in APR. LLMs take infilling templates with different qualities as input, and low-quality templates may misguide LLMs in constantly generating incorrect patches. Additionally, LLMs lack project-specific knowledge and struggle to leverage bug context fully. Therefore, we propose EvoAPR, which integrates the genetic algorithm and dynamic LoRA technique to enhance LLMs for better APR. First, to generate high-quality infilling templates, buggy codes are encoded to genetic representations, and genetic operators and multi-angle evaluation are designed to produce better templates. Then, to effectively utilize bug context, a bug-context-aware LoRA fine-tuning method is proposed to fuse various bug contexts by dynamically activating certain blocks of the LoRA module according to the bug index. Finally, these high-quality infilling templates are fed the fine-tuned LLMs for patch generation. Experimental results demonstrate that EvoAPR significantly improves LLMs' performance, surpassing state-of-the-art methods.

In large software projects with collaborative development, comprehensive code comments are crucial for code readability and maintainability. Code comments mainly include method comments and inline comments, where the former describes the functionality globally, and the latter describes the implementation details locally. Existing methods typically generate these two kinds of comments with specific locations independently, which results in weak correlations between comments and code context, as well as high model inference costs due to long token inputs. To address these issues, we define the combination of inline comments and method comments as Dual-Level Code Comments. We formulate the novel task of automatically generate dual-level code comments based on given code and propose an approach named DLCoG (DualLevel Code Comment Generation) to automate this task. First, a Semantic Segmentation and Identification multi-task model based on CodeBERT, termed Se2Iden (Semantic Segmentation and Identification model), is proposed to identify code segments requiring inline comments. Next, we retrieve similar samples to adopting the in-context learning paradigm, which can enhance the generation quality of large language models (LLMs) in specific domains. Finally, the LLM is guided to generate duallevel code comments using Chain-of-Thought (CoT) prompts that first produce inline comments, followed by method comments. We manually constructed a high-quality clean Java dataset consisting of*> based on open-source Java projects by (i) determining comments type and (ii) manually associating inline comments with their corresponding code. Then, we trained a multi-task learning model based on CodeBERT to automatically take the two steps needed, termed ICSA (Inline Comment Classification and Scope Association), thus to expand to a dataset containing 80k dual-level code comments. Experimental results on clean and extended datasets show that DLCoG outperforms all baselines by substantial margins. The contextual information provided by DLCoG can effectively improve the inline comments generated by LLM. Coordinated generation of dual-level comment also brings effective improvements to method comments, which is particularly significant when there are few contextual examples. Our work fills the long-standing gap in the dual-level code comment generation field, and can provide insights for future research in this direction. We provide open-source datasets and source code for future research.

In the era where Web3.0 values data security and privacy, adopting groundbreaking methods to enhance privacy in recommender systems is crucial. Recommender systems need to balance privacy and accuracy, while also having the ability to overcome cold start problems. The Differential Trust Mechanism (DTM) introduced in this paper is such an approach. The DTM provides a unique use of Gaussian distributions in modeling trust relationships within data, offering a novel way to balance recommendation accuracy with user privacy. This mechanism innovatively applies differential privacy principles, using Gaussian noise addition to protect individual user data from inference attacks, while maintaining the integrity and utility of the overall dataset. Unlike traditional anonymization techniques that often compromise data utility or vulnerability to reverse engineering, DTM provides a robust solution by dynamically adjusting privacy levels based on the trustworthiness of data requests. By combining DTM with existing mainstream recommendation algorithms, the prediction accuracy of MAE and RMSE increases by at least 6.60% and 2.69%, respectively. This dual benefit positions DTM as a significant advancement in secure data processing, especially relevant for online businesses and platforms where personalized recommendations are crucial yet privacy concerns are paramount.