arXiv subjects

Heng Li

Publications and source records attributed to Heng Li.

At least 91 records · Page 5Linked to original sources

arXiv2025

SeePhys: Does Seeing Help Thinking? -- Benchmarking Vision-Based Physics Reasoning

We present SeePhys, a large-scale multimodal benchmark for LLM reasoning grounded in physics questions ranging from middle school to PhD qualifying exams. The benchmark covers 7 fundamental domains spanning the physics discipline, incorporating 21 categories of highly heterogeneous diagrams. In contrast to prior works where visual elements mainly serve auxiliary purposes, our benchmark features a substantial proportion of vision-essential problems (75%) that mandate visual information extraction for correct solutions. Through extensive evaluation, we observe that even the most advanced visual reasoning models (e.g., Gemini-2.5-pro and o4-mini) achieve sub-60% accuracy on our benchmark. These results reveal fundamental challenges in current large language models' visual understanding capabilities, particularly in: (i) establishing rigorous coupling between diagram interpretation and physics reasoning, and (ii) overcoming their persistent reliance on textual cues as cognitive shortcuts.

cs.AI↗

arXiv2025

Challenges in structural variant calling in low-complexity regions

Background: Structural variants (SVs) are genomic differences $\ge$50 bp in length. They remain challenging to detect even with long sequence reads, and the sources of these difficulties are not well quantified. Results: We identified 35.4 Mb of low-complexity regions (LCRs) in GRCh38. Although these regions cover only 1.2% of the genome, they contain 69.1% of confident SVs in sample HG002. Across long-read SV callers, 77.3-91.3% of erroneous SV calls occur within LCRs, with error rates increasing with LCR length. Conclusion: SVs are enriched and difficult to call in LCRs. Special care need to be taken for calling and analyzing these variants.

q-bio.GN↗

arXiv2025

Improving spliced alignment by modeling splice sites with deep learning

Motivation: Spliced alignment refers to the alignment of messenger RNA (mRNA) or protein sequences to eukaryotic genomes. It plays a critical role in gene annotation and the study of gene functions. Accurate spliced alignment demands sophisticated modeling of splice sites, but current aligners use simple models, which may affect their accuracy given dissimilar sequences. Results: We implemented minisplice to learn splice signals with a one-dimensional convolutional neural network (1D-CNN) and trained a model with 7,026 parameters for vertebrate and insect genomes. It captures conserved splice signals across phyla and reveals GC-rich introns specific to mammals and birds. We used this model to estimate the empirical splicing probability for every GT and AG in genomes, and modified minimap2 and miniprot to leverage pre-computed splicing probability during alignment. Evaluation on human long-read RNA-seq data and cross-species protein datasets showed our method greatly improves the junction accuracy especially for noisy long RNA-seq reads and proteins of distant homology. Availability and implementation: https://github.com/lh3/minisplice

q-bio.GN↗

arXiv2025

Logging Requirement for Continuous Auditing of Responsible Machine Learning-based Applications

Machine learning (ML) is increasingly applied across industries to automate decision-making, but concerns about ethical and legal compliance remain due to limited transparency, fairness, and accountability. Monitoring through logging a long-standing practice in traditional software offers a potential means for auditing ML applications, as logs provide traceable records of system behavior useful for debugging, performance analysis, and continuous auditing. systematically auditing models for compliance or accountability. The findings underscore the need for enhanced logging practices and tooling that systematically integrate responsible AI metrics. Such practices would support the development of auditable, transparent, and ethically responsible ML systems, aligning with growing regulatory requirements and societal expectations. By highlighting specific deficiencies and opportunities, this work provides actionable guidance for both practitioners and tool developers seeking to strengthen the accountability and trustworthiness of ML applications.

cs.SE↗

arXiv2025

SAIL-Recon: Large SfM by Augmenting Scene Regression with Localization

Scene regression methods, such as VGGT, solve the Structure-from-Motion (SfM) problem by directly regressing camera poses and 3D scene structures from input images. They demonstrate impressive performance in handling images under extreme viewpoint changes. However, these methods struggle to handle a large number of input images. To address this problem, we introduce SAIL-Recon, a feed-forward Transformer for large scale SfM, by augmenting the scene regression network with visual localization capabilities. Specifically, our method first computes a neural scene representation from a subset of anchor images. The regression network is then fine-tuned to reconstruct all input images conditioned on this neural scene representation. Comprehensive experiments show that our method not only scales efficiently to large-scale scenes, but also achieves state-of-the-art results on both camera pose estimation and novel view synthesis benchmarks, including TUM-RGBD, CO3Dv2, and Tanks & Temples. We will publish our model and code. Code and models are publicly available at: https://hkust-sail.github.io/ sail-recon/.

cs.CV↗

arXiv2025

Plug it and Play on Logs: A Configuration-Free Statistic-Based Log Parser

Log parsing is an essential task in log analysis, and many tools have been designed to accomplish it. Existing log parsers can be categorized into statistic-based and semantic-based approaches. In comparison to semantic-based parsers, existing statistic-based parsers tend to be more efficient, require lower computational costs, and be more privacy-preserving thanks to on-premise deployment, but often fall short in their accuracy (e.g., grouping or parsing accuracy) and generalizability. Therefore, it became a common belief that statistic-based parsers cannot be as effective as semantic-based parsers since the latter could take advantage of external knowledge supported by pretrained language models. Our work, however, challenges this belief with a novel statistic-based parser, PIPLUP. PIPLUP eliminates the pre-assumption of the position of constant tokens for log grouping and relies on data-insensitive parameters to overcome the generalizability challenge, allowing "plug and play" on given log files. According to our experiments on an open-sourced large log dataset, PIPLUP shows promising accuracy and generalizability with the data-insensitive default parameter set. PIPLUP not only outperforms the state-of-the-art statistic-based log parsers, Drain and its variants, but also obtains a competitive performance compared to the best unsupervised semantic-based log parser (i.e., LUNAR). Further, PIPLUP exhibits low time consumption without GPU acceleration and external API usage; our simple, efficient, and effective approach makes it more practical in real-world adoptions, especially when costs and privacy are of major concerns.

cs.SE↗

arXiv2025

An Empirical Study on Method-Level Performance Evolution in Open-Source Java Projects

Performance is a critical quality attribute in software development, yet the impact of method-level code changes on performance evolution remains poorly understood. While developers often make intuitive assumptions about which types of modifications are likely to cause performance regressions or improvements, these beliefs lack empirical validation at a fine-grained level. We conducted a large-scale empirical study analyzing performance evolution in 15 mature open-source Java projects hosted on GitHub. Our analysis encompassed 739 commits containing 1,499 method-level code changes, using Java Microbenchmark Harness (JMH) for precise performance measurement and rigorous statistical analysis to quantify both the significance and magnitude of performance variations. We employed bytecode instrumentation to capture method-specific execution metrics and systematically analyzed four key aspects: temporal performance patterns, code change type correlations, developer and complexity factors, and domain-size interactions. Our findings reveal that 32.7% of method-level changes result in measurable performance impacts, with regressions occurring 1.3 times more frequently than improvements. Contrary to conventional wisdom, we found no significant differences in performance impact distributions across code change categories, challenging risk-stratified development strategies. Algorithmic changes demonstrate the highest improvement potential but carry substantial regression risk. Senior developers produce more stable changes with fewer extreme variations, while code complexity correlates with increased regression likelihood. Domain-size interactions reveal significant patterns, with web server + small projects exhibiting the highest performance instability. Our study provides empirical evidence for integrating automated performance testing into continuous integration pipelines.

cs.SE↗

arXiv2025

Finding easy regions for short-read variant calling from pangenome data

Background: While benchmarks on short-read variant calling suggest low error rate below 0.5%, they are only applicable to predefined confident regions. For a human sample without such regions, the error rate could be 10 times higher. Although multiple sets of easy regions have been identified to alleviate the issue, they fail to consider non-reference samples or are biased towards existing short-read data or aligners. Results: Here, using hundreds of high-quality human assemblies, we derived a set of sample-agnostic easy regions where short-read variant calling reaches high accuracy. These regions cover 88.2% of GRCh38, 92.2% of coding regions and 96.3% of ClinVar pathogenic variants. They achieve a good balance between coverage and easiness and can be generated for other human assemblies or species with multiple well assembled genomes. Conclusion: This resource provides a convient and powerful way to filter spurious variant calls for clinical or research human samples.

q-bio.GN↗

arXiv2025

From Technical Excellence to Practical Adoption: Lessons Learned Building an ML-Enhanced Trace Analysis Tool

System tracing has become essential for understanding complex software behavior in modern systems, yet sophisticated trace analysis tools face significant adoption gaps in industrial settings. Through a year-long collaboration with Ericsson Montréal, developing TMLL (Trace-Server Machine Learning Library, now in the Eclipse Foundation), we investigated barriers to trace analysis adoption. Contrary to assumptions about complexity or automation needs, practitioners struggled with translating expert knowledge into actionable insights, integrating analysis into their workflows, and trusting automated results they could not validate. We identified what we called the Excellence Paradox: technical excellence can actively impede adoption when conflicting with usability, transparency, and practitioner trust. TMLL addresses this through adoption-focused design that embeds expert knowledge in interfaces, provides transparent explanations, and enables incremental adoption. Validation through Ericsson's experts' feedback, Eclipse Foundation's integration, and a survey of 40 industry and academic professionals revealed consistent patterns: survey results showed that 77.5% prioritize quality and trust in results over technical sophistication, while 67.5% prefer semi-automated analysis with user control, findings supported by qualitative feedback from industrial collaboration and external peer review. Results validate three core principles: cognitive compatibility, embedded expertise, and transparency-based trust. This challenges conventional capability-focused tool development, demonstrating that sustainable adoption requires reorientation toward adoption-focused design with actionable implications for automated software engineering tools.

cs.SE↗

arXiv2025

PanTS: The Pancreatic Tumor Segmentation Dataset

PanTS is a large-scale, multi-institutional dataset curated to advance research in pancreatic CT analysis. It contains 36,390 CT scans from 145 medical centers, with expert-validated, voxel-wise annotations of over 993,000 anatomical structures, covering pancreatic tumors, pancreas head, body, and tail, and 24 surrounding anatomical structures such as vascular/skeletal structures and abdominal/thoracic organs. Each scan includes metadata such as patient age, sex, diagnosis, contrast phase, in-plane spacing, slice thickness, etc. AI models trained on PanTS achieve significantly better performance in pancreatic tumor detection, localization, and segmentation compared to those trained on existing public datasets. Our analysis indicates that these gains are directly attributable to the 16x larger-scale tumor annotations and indirectly supported by the 24 additional surrounding anatomical structures. As the largest and most comprehensive resource of its kind, PanTS offers a new benchmark for developing and evaluating AI models in pancreatic CT analysis.

eess.IV↗

arXiv2025

Unleashing Diffusion and State Space Models for Medical Image Segmentation

Existing segmentation models trained on a single medical imaging dataset often lack robustness when encountering unseen organs or tumors. Developing a robust model capable of identifying rare or novel tumor categories not present during training is crucial for advancing medical imaging applications. We propose DSM, a novel framework that leverages diffusion and state space models to segment unseen tumor categories beyond the training data. DSM utilizes two sets of object queries trained within modified attention decoders to enhance classification accuracy. Initially, the model learns organ queries using an object-aware feature grouping strategy to capture organ-level visual features. It then refines tumor queries by focusing on diffusion-based visual prompts, enabling precise segmentation of previously unseen tumors. Furthermore, we incorporate diffusion-guided feature fusion to improve semantic segmentation performance. By integrating CLIP text embeddings, DSM captures category-sensitive classes to improve linguistic transfer knowledge, thereby enhancing the model's robustness across diverse scenarios and multi-label tasks. Extensive experiments demonstrate the superior performance of DSM in various tumor segmentation tasks. Code is available at https://github.com/Rows21/k-Means_Mask_Mamba.

cs.CV↗

arXiv2025

Adversarial Attack Classification and Robustness Testing for Large Language Models for Code

Large Language Models (LLMs) have become vital tools in software development tasks such as code generation, completion, and analysis. As their integration into workflows deepens, ensuring robustness against vulnerabilities especially those triggered by diverse or adversarial inputs becomes increasingly important. Such vulnerabilities may lead to incorrect or insecure code generation when models encounter perturbed task descriptions, code, or comments. Prior research often overlooks the role of natural language in guiding code tasks. This study investigates how adversarial perturbations in natural language inputs including prompts, comments, and descriptions affect LLMs for Code (LLM4Code). It examines the effects of perturbations at the character, word, and sentence levels to identify the most impactful vulnerabilities. We analyzed multiple projects (e.g., ReCode, OpenAttack) and datasets (e.g., HumanEval, MBPP), establishing a taxonomy of adversarial attacks. The first dimension classifies the input type code, prompts, or comments while the second dimension focuses on granularity: character, word, or sentence-level changes. We adopted a mixed-methods approach, combining quantitative performance metrics with qualitative vulnerability analysis. LLM4Code models show varying robustness across perturbation types. Sentence-level attacks were least effective, suggesting models are resilient to broader contextual changes. In contrast, word-level perturbations posed serious challenges, exposing semantic vulnerabilities. Character-level effects varied, showing model sensitivity to subtle syntactic deviations.Our study offers a structured framework for testing LLM4Code robustness and emphasizes the critical role of natural language in adversarial evaluation. Improving model resilience to semantic-level disruptions is essential for secure and reliable code-generation systems.

cs.SE↗

arXiv2025

RADAR: Enhancing Radiology Report Generation with Supplementary Knowledge Injection

Large language models (LLMs) have demonstrated remarkable capabilities in various domains, including radiology report generation. Previous approaches have attempted to utilize multimodal LLMs for this task, enhancing their performance through the integration of domain-specific knowledge retrieval. However, these approaches often overlook the knowledge already embedded within the LLMs, leading to redundant information integration. To address this limitation, we propose Radar, a framework for enhancing radiology report generation with supplementary knowledge injection. Radar improves report generation by systematically leveraging both the internal knowledge of an LLM and externally retrieved information. Specifically, it first extracts the model's acquired knowledge that aligns with expert image-based classification outputs. It then retrieves relevant supplementary knowledge to further enrich this information. Finally, by aggregating both sources, Radar generates more accurate and informative radiology reports. Extensive experiments on MIMIC-CXR, CheXpert-Plus, and IU X-ray demonstrate that our model outperforms state-of-the-art LLMs in both language quality and clinical accuracy.

cs.CV↗

arXiv2025

SDLog: A Deep Learning Framework for Detecting Sensitive Information in Software Logs

Software logs are messages recorded during the execution of a software system that provide crucial run-time information about events and activities. Although software logs have a critical role in software maintenance and operation tasks, publicly accessible log datasets remain limited, hindering advance in log analysis research and practices. The presence of sensitive information, particularly Personally Identifiable Information (PII) and quasi-identifiers, introduces serious privacy and re-identification risks, discouraging the publishing and sharing of real-world logs. In practice, log anonymization techniques primarily rely on regular expression patterns, which involve manually crafting rules to identify and replace sensitive information. However, these regex-based approaches suffer from significant limitations, such as extensive manual efforts and poor generalizability across diverse log formats and datasets. To mitigate these limitations, we introduce SDLog, a deep learning-based framework designed to identify sensitive information in software logs. Our results show that SDLog overcomes regex limitations and outperforms the best-performing regex patterns in identifying sensitive information. With only 100 fine-tuning samples from the target dataset, SDLog can correctly identify 99.5% of sensitive attributes and achieves an F1-score of 98.4%. To the best of our knowledge, this is the first deep learning alternative to regex-based methods in software log anonymization.

cs.SE↗

arXiv2025

An Efficient Model Maintenance Approach for MLOps

In recent years, many industries have utilized machine learning (ML) models in their systems. Ideally, ML models should be trained on and applied to data from the same distributions. However, the data evolves over time in many application areas, leading to concept drift, which in turn causes the performance of the ML models to degrade over time. Therefore, maintaining up-to-date ML models plays a critical role in the MLOps pipeline. Existing ML model maintenance approaches are often computationally resource-intensive, costly, time-consuming, and model-dependent. Thus, we propose an improved MLOps pipeline, a new model maintenance approach and a Similarity-Based Model Reuse (SimReuse) tool to address the challenges of ML model maintenance. We identify seasonal and recurrent data distribution patterns in time series datasets throughout a preliminary study. Recurrent data distribution patterns enable us to reuse previously trained models for similar distributions in the future, thus avoiding frequent unnecessary retrainings. Then, we integrated the model reuse approach into the MLOps pipeline and proposed our improved MLOps pipeline. Furthermore, we develop SimReuse, a tool to implement the new components of our MLOps pipeline to store models and reuse them for inference of data segments with similar data distributions in the future. Our evaluation results on five time series datasets demonstrate that our model reuse approach can maintain the models' performance while significantly reducing maintenance time, costs, and the number of retrainings. Our model reuse approach achieves ML model performance comparable to the best baselines, while reducing the computation time and costs to 1/8th. Therefore, industries and practitioners can benefit from our approach and use our tool to maintain their ML models' performance in the deployment phase to reduce their maintenance time and costs.

cs.SE↗

arXiv2025

Document Image Rectification Bases on Self-Adaptive Multitask Fusion

Deformed document image rectification is essential for real-world document understanding tasks, such as layout analysis and text recognition. However, current multi-task methods -- such as background removal, 3D coordinate prediction, and text line segmentation -- often overlook the complementary features between tasks and their interactions. To address this gap, we propose a self-adaptive learnable multi-task fusion rectification network named SalmRec. This network incorporates an inter-task feature aggregation module that adaptively improves the perception of geometric distortions, enhances feature complementarity, and reduces negative interference. We also introduce a gating mechanism to balance features both within global tasks and between local tasks effectively. Experimental results on two English benchmarks (DIR300 and DocUNet) and one Chinese benchmark (DocReal) demonstrate that our method significantly improves rectification performance. Ablation studies further highlight the positive impact of different tasks on dewarping and the effectiveness of our proposed module.

cs.CV↗

arXiv2025

Protecting Privacy in Software Logs: What Should Be Anonymized?

Software logs, generated during the runtime of software systems, are essential for various development and analysis activities, such as anomaly detection and failure diagnosis. However, the presence of sensitive information in these logs poses significant privacy concerns, particularly regarding Personally Identifiable Information (PII) and quasi-identifiers that could lead to re-identification risks. While general data privacy has been extensively studied, the specific domain of privacy in software logs remains underexplored, with inconsistent definitions of sensitivity and a lack of standardized guidelines for anonymization. To mitigate this gap, this study offers a comprehensive analysis of privacy in software logs from multiple perspectives. We start by performing an analysis of 25 publicly available log datasets to identify potentially sensitive attributes. Based on the result of this step, we focus on three perspectives: privacy regulations, research literature, and industry practices. We first analyze key data privacy regulations, such as the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA), to understand the legal requirements concerning sensitive information in logs. Second, we conduct a systematic literature review to identify common privacy attributes and practices in log anonymization, revealing gaps in existing approaches. Finally, we survey 45 industry professionals to capture practical insights on log anonymization practices. Our findings shed light on various perspectives of log privacy and reveal industry challenges, such as technical and efficiency issues while highlighting the need for standardized guidelines. By combining insights from regulatory, academic, and industry perspectives, our study aims to provide a clearer framework for identifying and protecting sensitive information in software logs.

cs.SE↗

arXiv2025

Can We Recycle Our Old Models? An Empirical Evaluation of Model Selection Mechanisms for AIOps Solutions

AIOps (Artificial Intelligence for IT Operations) solutions leverage the tremendous amount of data produced during the operation of large-scale systems and machine learning models to assist software practitioners in their system operations. Existing AIOps solutions usually maintain AIOps models against concept drift through periodical retraining, despite leaving a pile of discarded historical models that may perform well on specific future data. Other prior works propose dynamically selecting models for prediction tasks from a set of candidate models to optimize the model performance. However, there is no prior work in the AIOps area that assesses the use of model selection mechanisms on historical models to improve model performance or robustness. To fill the gap, we evaluate several model selection mechanisms by assessing their capabilities in selecting the optimal AIOps models that were built in the past to make predictions for the target data. We performed a case study on three large-scale public operation datasets: two trace datasets from the cloud computing platforms of Google and Alibaba, and one disk stats dataset from the BackBlaze cloud storage data center. We observe that the model selection mechnisms utilizing temporal adjacency tend to have a better performance and can prevail the periodical retraining approach. Our findings also highlight a performance gap between existing model selection mechnisms and the theoretical upper bound which may motivate future researchers and practitioners in investigating more efficient and effective model selection mechanisms that fit in the context of AIOps.

cs.SE↗