Jacky Keung

Fengji Zhang, Jin Liu, Yao Wan et al.

Stack Overflow is one of the most popular programming communities where developers can seek help for their encountered problems. Nevertheless, if inexperienced developers fail to describe their problems clearly, it is hard for them to attract sufficient attention and get the anticipated answers. We propose M$_3$NSCT5, a novel approach to automatically generate multiple post titles from the given code snippets. Developers may use the generated titles to find closely related posts and complete their problem descriptions. M$_3$NSCT5 employs the CodeT5 backbone, which is a pre-trained Transformer model having an excellent language understanding and generation ability. To alleviate the ambiguity issue that the same code snippets could be aligned with different titles under varying contexts, we propose the maximal marginal multiple nucleus sampling strategy to generate multiple high-quality and diverse title candidates at a time for the developers to choose from. We build a large-scale dataset with 890,000 question posts covering eight programming languages to validate the effectiveness of M$_3$NSCT5. The automatic evaluation results on the BLEU and ROUGE metrics demonstrate the superiority of M$_3$NSCT5 over six state-of-the-art baseline models. Moreover, a human evaluation with trustworthy results also demonstrates the great potential of our approach for real-world application.

Wenqiang Luo, Jacky Wai Keung, Boyang Yang et al.

Recent advances in leveraging LLMs for APR have demonstrated impressive capabilities in fixing software defects. However, current LLM-based approaches predominantly focus on mainstream programming languages like Java and Python, neglecting less prevalent but emerging languages such as Rust due to expensive training resources, limited datasets, and insufficient community support. This narrow focus creates a significant gap in repair capabilities across the programming language spectrum, where the full potential of LLMs for comprehensive multilingual program repair remains largely unexplored. To address this limitation, we introduce a novel cross-language program repair approach LANTERN that leverages LLMs' differential proficiency across languages through a multi-agent iterative repair paradigm. Our technique strategically translates defective code from languages where LLMs exhibit weaker repair capabilities to languages where they demonstrate stronger performance, without requiring additional training. A key innovation of our approach is an LLM-based decision-making system that dynamically selects optimal target languages based on bug characteristics and continuously incorporates feedback from previous repair attempts. We evaluate our method on xCodeEval, a comprehensive multilingual benchmark comprising 5,068 bugs across 11 programming languages. Results demonstrate significant enhancement in repair effectiveness, particularly for underrepresented languages, with Rust showing a 22.09% improvement in Pass@10 metrics. Our research provides the first empirical evidence that cross-language translation significantly expands the repair capabilities of LLMs and effectively bridges the performance gap between programming languages with different levels of popularity, opening new avenues for truly language-agnostic automated program repair.

Yuchen Cao, Hanlin Zhang, Jacky Wai Keung et al.

Large language models (LLMs) are increasingly used as quantitative research copilots to translate natural-language strategy specifications into executable trading code. Yet most existing evaluations either focus on static financial knowledge or summarize performance with a single profitability metric, leaving a gap for benchmarking strategy-to-code trading systems as governed, auditable software. We introduce SysTradeBench (SysTB), an iterative build-test-patch benchmark that evaluates LLM-generated trading systems under drift-aware diagnostics. Given a standardized Base Strategy Doc and frozen semantics, each model must produce (i) a strategy card, (ii) executable code, and (iii) mandatory audit logs. A sandboxed harness runs determinism and anti-leakage checks, detects rule drift across iterations, and returns evidence bundles to support constrained patches. SysTradeBench reports multi-dimensional scorecards for spec fidelity, risk discipline, reliability, and out-of-sample robustness indicators, together with cost-effectiveness signals. We evaluate 17 models across 12 strategies. Top models achieve validity above 91.7 percent with strong aggregate scores, but evidence-driven iteration also induces code convergence by Iter2. These findings suggest that LLM iteration complements rather than replaces human quantitative researcher governance: LLMs excel at rapid prototyping and shallow bug fixes, while human oversight remains essential for critical strategies requiring solution diversity and ensemble robustness.

Linquan Wu, Tianxiang Jiang, Yifei Dong et al.

Current multimodal latent reasoning often relies on external supervision (e.g., auxiliary images), ignoring intrinsic visual attention dynamics. In this work, we identify a critical Perception Gap in distillation: student models frequently mimic a teacher's textual output while attending to fundamentally divergent visual regions, effectively relying on language priors rather than grounded perception. To bridge this, we propose LaViT, a framework that aligns latent visual thoughts rather than static embeddings. LaViT compels the student to autoregressively reconstruct the teacher's visual semantics and attention trajectories prior to text generation, employing a curriculum sensory gating mechanism to prevent shortcut learning. Extensive experiments show that LaViT significantly enhances visual grounding, achieving up to +16.9% gains on complex reasoning tasks and enabling a compact 3B model to outperform larger open-source variants and proprietary models like GPT-4o.

Fengji Zhang, Linquan Wu, Huiyu Bai et al.

Understanding and reasoning over diagrams is a fundamental aspect of human intelligence. While Large Multimodal Models (LMMs) have demonstrated impressive capabilities across various tasks, existing benchmarks lack comprehensive evaluation of their diagram interpretation and reasoning abilities, particularly in coding contexts. We present HumanEval-V, a rigorous benchmark of human-annotated coding tasks that spans six task types and evaluates diverse visual reasoning capabilities. Each task features carefully crafted diagrams paired with function signatures and test cases, employing novel code generation tasks to thoroughly assess models' diagram comprehension. Through extensive experiments with 22 LMMs, we find that even top-performing models achieve modest success rates, with Claude 3.5 Sonnet reaching only 36.8% pass@1, highlighting substantial room for improvement. Our analysis reveals that current LMMs struggle with spatial transformations, topological relationships, and dynamic patterns that humans find intuitive. These findings provide valuable insights for advancing LMMs' visual reasoning abilities. We have open-sourced our code and benchmark at https://github.com/HumanEval-V/HumanEval-V-Benchmark.

Zhen Yang, Fang Liu, Zhongxing Yu et al.

Code translation tools (transpilers) are developed for automatic source-to-source translation. Although learning-based transpilers have shown impressive enhancement against rule-based counterparts, owing to their task-specific pre-training on extensive monolingual corpora. Their current performance still remains unsatisfactory for practical deployment, and the associated training resources are also prohibitively expensive. LLMs pre-trained on huge amounts of human-written code/text have shown remarkable performance in many code intelligence tasks due to their powerful generality, even without task-specific training. Thus, LLMs can potentially circumvent the above limitations, but they have not been exhaustively explored yet. This paper investigates diverse LLMs and learning-based transpilers for automated code translation tasks, finding that: although certain LLMs have outperformed current transpilers, they still have some accuracy issues, where most of the failures are induced by a lack of comprehension of source programs, missing clear instructions on I/O types in translation, and ignoring discrepancies between source and target programs. Enlightened by the above findings, we further propose UniTrans, a Unified code Translation framework, applicable to various LLMs, for unleashing their power in this field. Specifically, UniTrans first crafts a series of test cases for target programs with the assistance of source programs. Next, it harnesses the above auto-generated test cases to augment the code translation and then evaluate their correctness via execution. Afterward, UniTrans further (iteratively) repairs incorrectly translated programs prompted by test case execution results. Extensive experiments are conducted on six settings of translation datasets between Python, Java, and C++. Three recent LLMs of diverse sizes are tested with UniTrans, and all achieve substantial improvements.

Fengji Zhang, Xinyao Niu, Chengyang Ying et al. · tsinghua

Recent advances in Large Language Models (LLMs) and Reinforcement Learning (RL) have led to strong performance in open-domain question answering (QA). However, existing models still struggle with questions that admit multiple valid answers. Standard QA benchmarks, which typically assume a single gold answer, overlook this reality and thus produce inappropriate training signals. Existing attempts to handle ambiguity often rely on costly manual annotation, which is difficult to scale to multi-hop datasets such as HotpotQA and MuSiQue. In this paper, we present A$^2$Search, an annotation-free, end-to-end training framework to recognize and handle ambiguity. At its core is an automated pipeline that detects ambiguous questions and gathers alternative answers via trajectory sampling and evidence verification. The model is then optimized with RL using a carefully designed $\mathrm{AnsF1}$ reward, which naturally accommodates multiple answers. Experiments on eight open-domain QA benchmarks demonstrate that A$^2$Search achieves new state-of-the-art performance. With only a single rollout, A$^2$Search-7B yields an average $\mathrm{AnsF1}@1$ score of $48.4\%$ across four multi-hop benchmarks, outperforming all strong baselines, including the substantially larger ReSearch-32B ($46.2\%$). Extensive analyses further show that A$^2$Search resolves ambiguity and generalizes across benchmarks, highlighting that embracing ambiguity is essential for building more reliable QA systems. Our code, data, and model weights can be found at https://github.com/zfj1998/A2Search

Xinrui He, Shuo Liu, Jackey Keung et al.

As data privacy and security attract increasing attention, Federated Recommender System (FRS) offers a solution that strikes a balance between providing high-quality recommendations and preserving user privacy. However, the presence of statistical heterogeneity in FRS, commonly observed due to personalized decision-making patterns, can pose challenges. To address this issue and maximize the benefit of collaborative filtering (CF) in FRS, it is intuitive to consider clustering clients (users) as well as items into different groups and learning group-specific models. Existing methods either resort to client clustering via user representations-risking privacy leakage, or employ classical clustering strategies on item embeddings or gradients, which we found are plagued by the curse of dimensionality. In this paper, we delve into the inefficiencies of the K-Means method in client grouping, attributing failures due to the high dimensionality as well as data sparsity occurring in FRS, and propose CoFedRec, a novel Co-clustering Federated Recommendation mechanism, to address clients heterogeneity and enhance the collaborative filtering within the federated framework. Specifically, the server initially formulates an item membership from the client-provided item networks. Subsequently, clients are grouped regarding a specific item category picked from the item membership during each communication round, resulting in an intelligently aggregated group model. Meanwhile, to comprehensively capture the global inter-relationships among items, we incorporate an additional supervised contrastive learning term based on the server-side generated item membership into the local training phase for each client. Extensive experiments on four datasets are provided, which verify the effectiveness of the proposed CoFedRec.

Linquan Wu, Tianxiang Jiang, Haoyu Yang et al.

Micro-expression recognition (MER) has achieved impressive accuracy in controlled laboratory settings. However, its real-world applicability faces a significant generalization cliff, severely hindering practical deployment due to poor performance on unseen data and susceptibility to domain shifts. Existing attention mechanisms often overfit to dataset-specific appearance cues or rely on fixed spatial priors, making them fragile in diverse environments. We posit that robust MER requires focusing on quasi-invariant motion orientations inherent to micro-expressions, rather than superficial pixel-level features. To this end, we introduce \textbf{FaceSleuth-R}, a framework centered on our novel \textbf{Single-Orientation Attention (SOA)} module. SOA is a lightweight, differentiable operator that enables the network to learn layer-specific optimal orientations, effectively guiding attention towards these robust motion cues. Through extensive experiments, we demonstrate that SOA consistently discovers a universal near-vertical motion prior across diverse datasets. More critically, FaceSleuth-R showcases superior generalization in rigorous Leave-One-Dataset-Out (LODO) protocols, significantly outperforming baselines and state-of-the-art methods when confronted with domain shifts. Furthermore, our approach establishes \textbf{state-of-the-art results} across several benchmarks. This work highlights adaptive orientation-aware attention as a key paradigm for developing truly generalized and high-performing MER systems.

Fengji Zhang, Xiao Yu, Jacky Keung et al.

Context: Stack Overflow is very helpful for software developers who are seeking answers to programming problems. Previous studies have shown that a growing number of questions are of low quality and thus obtain less attention from potential answerers. Gao et al. proposed an LSTM-based model (i.e., BiLSTM-CC) to automatically generate question titles from the code snippets to improve the question quality. However, only using the code snippets in the question body cannot provide sufficient information for title generation, and LSTMs cannot capture the long-range dependencies between tokens. Objective: This paper proposes CCBERT, a deep learning based novel model to enhance the performance of question title generation by making full use of the bi-modal information of the entire question body. Method: CCBERT follows the encoder-decoder paradigm and uses CodeBERT to encode the question body into hidden representations, a stacked Transformer decoder to generate predicted tokens, and an additional copy attention layer to refine the output distribution. Both the encoder and decoder perform the multi-head self-attention operation to better capture the long-range dependencies. This paper builds a dataset containing around 200,000 high-quality questions filtered from the data officially published by Stack Overflow to verify the effectiveness of the CCBERT model. Results: CCBERT outperforms all the baseline models on the dataset. Experiments on both code-only and low-resource datasets show the superiority of CCBERT with less performance degradation. The human evaluation also shows the excellent performance of CCBERT concerning both readability and correlation criteria.

Yutian Tang, Haoyu Wang, Xian Zhan et al.

Being able to automatically detect the performance issues in apps can significantly improve apps' quality as well as having a positive influence on user satisfaction. Application Performance Management (APM) libraries are used to locate the apps' performance bottleneck, monitor their behaviors at runtime, and identify potential security risks. Although app developers have been exploiting application performance management (APM) tools to capture these potential performance issues, most of them do not fully understand the internals of these APM tools and the effect on their apps. To fill this gap, in this paper, we conduct the first systematic study on APMs for apps by scrutinizing 25 widely-used APMs for Android apps and develop a framework named APMHunter for exploring the usage of APMs in Android apps. Using APMHunter, we conduct a large-scale empirical study on 500,000 Android apps to explore the usage patterns of APMs and discover the potential misuses of APMs. We obtain two major findings: 1) some APMs still employ deprecated permissions and approaches, which makes APMs fail to perform as expected; 2) inappropriate use of APMs can cause privacy leaks. Thus, our study suggests that both APM vendors and developers should design and use APMs scrupulously.

Shujuan Jiang, Miao Zhang, Yanmei Zhang et al.

Integration testing is a very important step in software testing. Existing methods evaluate the stubbing cost for class integration test orders by considering only the interclass direct relationships such as inheritance, aggregation, and association, but they omit the interclass indirect relationship caused by control coupling, which can also affect the test orders and the stubbing cost. In this paper, we introduce an integration test order strategy to consider control coupling. We advance the concept of transitive relationship to describe this kind of interclass dependency and propose a new measurement method to estimate the complexity of control coupling, which is the complexity of stubs created for a transitive relationship. We evaluate our integration test order strategy on 10 programs on various scales. The results show that considering the transitive relationship when generating class integration test orders can significantly reduce the stubbing cost for most programs and that our integration test order strategy obtains satisfactory results more quickly than other methods.

Zhen Yang, Jacky Keung, Xiao Yu et al.

Code comment has been an important part of computer programs, greatly facilitating the understanding and maintenance of source code. However, high-quality code comments are often unavailable in smart contracts, the increasingly popular programs that run on the blockchain. In this paper, we propose a Multi-Modal Transformer-based (MMTrans) code summarization approach for smart contracts. Specifically, the MMTrans learns the representation of source code from the two heterogeneous modalities of the Abstract Syntax Tree (AST), i.e., Structure-based Traversal (SBT) sequences and graphs. The SBT sequence provides the global semantic information of AST, while the graph convolution focuses on the local details. The MMTrans uses two encoders to extract both global and local semantic information from the two modalities respectively, and then uses a joint decoder to generate code comments. Both the encoders and the decoder employ the multi-head attention structure of the Transformer to enhance the ability to capture the long-range dependencies between code tokens. We build a dataset with over 300K <method, comment> pairs of smart contracts, and evaluate the MMTrans on it. The experimental results demonstrate that the MMTrans outperforms the state-of-the-art baselines in terms of four evaluation metrics by a substantial margin, and can generate higher quality comments.

Zheng Li, Liam O'Brien, Jacky Keung et al.

Within the service-oriented computing domain, Web service composition is an effective realization to satisfy the rapidly changing requirements of business. Therefore, the research into Web service composition has unfolded broadly. Since examining all of the related work in this area becomes a mission next to impossible, the classification of composition approaches can be used to facilitate multiple research tasks. However, the current attempts to classify Web service composition do not have clear objectives. Furthermore, the contexts and technologies of composition approaches are confused in the existing classifications. This paper proposes an effort-oriented classification matrix for Web service composition, which distinguishes between the context and technology dimension. The context dimension is aimed at analyzing the environment influence on the effort of Web service composition, while the technology dimension focuses on the technique influence on the effort. Consequently, besides the traditional classification benefits, this matrix can be used to build the basis of cost estimation for Web service composition in future research.

Zheng Li, Jacky Keung

Due to the complexity of Service-Oriented Architecture (SOA), cost and effort estimation for SOA-based software development is more difficult than that for traditional software development. Unfortunately, there is a lack of published work about cost and effort estimation for SOA-based software. Existing cost estimation approaches are inadequate to address the complex service-oriented systems. This paper proposes a novel framework based on Divide-and-Conquer (D&C) for cost estimation for building SOA-based software. By dealing with separately development parts, the D&C framework can help organizations simplify and regulate SOA implementation cost estimation. Furthermore, both cost estimation modeling and software sizing work can be satisfied respectively by switching the corresponding metrics within this framework. Given the requirement of developing these metrics, this framework also defines the future research in four different directions according to the separate cost estimation sub-problems.