Zihan Wu

Jie Xu, Zihan Wu, Cong Wang et al.

Machine learning models may inadvertently memorize sensitive, unauthorized, or malicious data, posing risks of privacy breaches, security vulnerabilities, and performance degradation. To address these issues, machine unlearning has emerged as a critical technique to selectively remove specific training data points' influence on trained models. This paper provides a comprehensive taxonomy and analysis of the solutions in machine unlearning. We categorize existing solutions into exact unlearning approaches that remove data influence thoroughly and approximate unlearning approaches that efficiently minimize data influence. By comprehensively reviewing solutions, we identify and discuss their strengths and limitations. Furthermore, we propose future directions to advance machine unlearning and establish it as an essential capability for trustworthy and adaptive machine learning models. This paper provides researchers with a roadmap of open problems, encouraging impactful contributions to address real-world needs for selective data removal.

Zihan Wu, Neil Scheidwasser-Clow, Karl El Hajal et al. · tencent-ai

Since the mental states of the speaker modulate speech, stress introduced by cognitive or physical loads could be detected in the voice. The existing voice stress detection benchmark has shown that the audio embeddings extracted from the Hybrid BYOL-S self-supervised model perform well. However, the benchmark only evaluates performance separately on each dataset, but does not evaluate performance across the different types of stress and different languages. Moreover, previous studies found strong individual differences in stress susceptibility. This paper presents the design and development of voice stress detection, trained on more than 100 speakers from 9 language groups and five different types of stress. We address individual variabilities in voice stress analysis by adding speaker embeddings to the hybrid BYOL-S features. The proposed method significantly improves voice stress detection performance with an input audio length of only 3-5 seconds.

Karl El Hajal, Zihan Wu, Neil Scheidwasser-Clow et al.

Automatic speech quality assessment is essential for audio researchers, developers, speech and language pathologists, and system quality engineers. The current state-of-the-art systems are based on framewise speech features (hand-engineered or learnable) combined with time dependency modeling. This paper proposes an efficient system with results comparable to the best performing model in the ConferencingSpeech 2022 challenge. Our proposed system is characterized by a smaller number of parameters (40-60x), fewer FLOPS (100x), lower memory consumption (10-15x), and lower latency (30x). Speech quality practitioners can therefore iterate much faster, deploy the system on resource-limited hardware, and, overall, the proposed system contributes to sustainable machine learning. The paper also concludes that framewise embeddings outperform utterance-level embeddings and that multi-task training with acoustic conditions modeling does not degrade speech quality prediction while providing better interpretation.

Jie Xu, Zihan Wu, Cong Wang et al.

To address the growing demand for privacy protection in machine learning, we propose a novel and efficient machine unlearning approach for \textbf{L}arge \textbf{M}odels, called \textbf{LM}Eraser. Existing unlearning research suffers from entangled training data and complex model architectures, incurring extremely high computational costs for large models. LMEraser takes a divide-and-conquer strategy with a prompt tuning architecture to isolate data influence. The training dataset is partitioned into public and private datasets. Public data are used to train the backbone of the model. Private data are adaptively clustered based on their diversity, and each cluster is used to optimize a prompt separately. This adaptive prompt tuning mechanism reduces unlearning costs and maintains model performance. Experiments demonstrate that LMEraser achieves a $100$-fold reduction in unlearning costs without compromising accuracy compared to prior work. Our code is available at: \url{https://github.com/lmeraser/lmeraser}.

Wentao Pan, Zhe Xu, Jiangpeng Yan et al. · tencent-ai

Semi-supervised semantic segmentation aims to utilize limited labeled images and abundant unlabeled images to achieve label-efficient learning, wherein the weak-to-strong consistency regularization framework, popularized by FixMatch, is widely used as a benchmark scheme. Despite its effectiveness, we observe that such scheme struggles with satisfactory segmentation for the local regions. This can be because it originally stems from the image classification task and lacks specialized mechanisms to capture fine-grained local semantics that prioritizes in dense prediction. To address this issue, we propose a novel framework called \texttt{MaskMatch}, which enables fine-grained locality learning to achieve better dense segmentation. On top of the original teacher-student framework, we design a masked modeling proxy task that encourages the student model to predict the segmentation given the unmasked image patches (even with 30\% only) and enforces the predictions to be consistent with pseudo-labels generated by the teacher model using the complete image. Such design is motivated by the intuition that if the predictions are more consistent given insufficient neighboring information, stronger fine-grained locality perception is achieved. Besides, recognizing the importance of reliable pseudo-labels in the above locality learning and the original consistency learning scheme, we design a multi-scale ensembling strategy that considers context at different levels of abstraction for pseudo-label generation. Extensive experiments on benchmark datasets demonstrate the superiority of our method against previous approaches and its plug-and-play flexibility.

Jiali Chen, Xusen Hei, Yuqi Xue et al.

Chinese literary classics hold significant cultural and educational value, offering deep insights into morality, history, and human nature. These works often include classical Chinese and complex narratives, making them difficult for children to read. To bridge this gap, we introduce a child-friendly literary adaptation (CLA) task to adapt the Chinese literary classic into engaging and accessible text for children. However, recent large language models (LLMs) overlook children's reading preferences (\ie, vivid character portrayals, concise narrative structures, and appropriate readability), which poses challenges in CLA. In this paper, we propose a method called InstructChild, which augments the LLM with these preferences for adaptation. Specifically, we first obtain the characters' personalities and narrative structure as additional information for fine-grained instruction tuning. Then, we devise a readability metric as the reward to align the LLM with the children's reading level. Finally, a lookahead decoding strategy is applied to improve the readability of the generated text during inference. To support the evaluation of CLA task, we construct the Classic4Children dataset, which comprises both the original and child-friendly versions of the Four Great Classical Novels of Chinese literature. Experimental results show that our InstructChild significantly improves automatic and human evaluation performance.

Yuhang Huang, Zihan Wu, Chongyang Gao et al.

Large Vision-Language Models (LVLMs) are gaining traction for their remarkable ability to process and integrate visual and textual data. Despite their popularity, the capacity of LVLMs to generate precise, fine-grained textual descriptions has not been fully explored. This study addresses this gap by focusing on \textit{distinctiveness} and \textit{fidelity}, assessing how models like Open-Flamingo, IDEFICS, and MiniGPT-4 can distinguish between similar objects and accurately describe visual features. We proposed the Textual Retrieval-Augmented Classification (TRAC) framework, which, by leveraging its generative capabilities, allows us to delve deeper into analyzing fine-grained visual description generation. This research provides valuable insights into the generation quality of LVLMs, enhancing the understanding of multimodal language models. Notably, MiniGPT-4 stands out for its better ability to generate fine-grained descriptions, outperforming the other two models in this aspect. The code is provided at \url{https://anonymous.4open.science/r/Explore_FGVDs-E277}.

Zihan Wu, Jie Xu, Yun Peng et al.

Large Language Models (LLMs) struggle to automate real-world vulnerability detection due to two key limitations: the heterogeneity of vulnerability patterns undermines the effectiveness of a single unified model, and manual prompt engineering for massive weakness categories is unscalable. To address these challenges, we propose \textbf{MulVul}, a retrieval-augmented multi-agent framework designed for precise and broad-coverage vulnerability detection. MulVul adopts a coarse-to-fine strategy: a \emph{Router} agent first predicts the top-$k$ coarse categories and then forwards the input to specialized \emph{Detector} agents, which identify the exact vulnerability types. Both agents are equipped with retrieval tools to actively source evidence from vulnerability knowledge bases to mitigate hallucinations. Crucially, to automate the generation of specialized prompts, we design \emph{Cross-Model Prompt Evolution}, a prompt optimization mechanism where a generator LLM iteratively refines candidate prompts while a distinct executor LLM validates their effectiveness. This decoupling mitigates the self-correction bias inherent in single-model optimization. Evaluated on 130 CWE types, MulVul achieves 34.79\% Macro-F1, outperforming the best baseline by 41.5\%. Ablation studies validate cross-model prompt evolution, which boosts performance by 51.6\% over manual prompts by effectively handling diverse vulnerability patterns.

Jiali Chen, Yujie Jia, Zihan Wu et al.

Experiment commentary is crucial in describing the experimental procedures, delving into underlying scientific principles, and incorporating content-related safety guidelines. In practice, human teachers rely heavily on subject-specific expertise and invest significant time preparing such commentary. To address this challenge, we introduce the task of automatic commentary generation across multi-discipline scientific experiments. While recent progress in large multimodal models (LMMs) has demonstrated promising capabilities in video understanding and reasoning, their ability to generate fine-grained and insightful experiment commentary remains largely underexplored. In this paper, we make the following contributions: (i) We construct \textit{ExpInstruct}, the first dataset tailored for experiment commentary generation, featuring over 7\textit{K} step-level commentaries across 21 scientific subjects from 3 core disciplines (\ie, science, healthcare and engineering). Each sample includes procedural descriptions along with potential scientific principles (\eg, chemical equations and physical laws) and safety guidelines. (ii) We propose ExpStar, an automatic experiment commentary generation model that leverages a retrieval-augmented mechanism to adaptively access, evaluate, and utilize external knowledge. (iii) Extensive experiments show that our ExpStar substantially outperforms 14 leading LMMs, which highlights the superiority of our dataset and model. We believe that ExpStar holds great potential for advancing AI-assisted scientific experiment instruction.

Zihan Wu, Yicheng Tang, Barbara Ericson

AI-supported tools can help learners overcome challenges in programming education by providing adaptive assistance. However, existing research often focuses on individual tools rather than deriving broader design recommendations. A key challenge in designing these systems is balancing learner control with system-driven guidance. To explore user preferences for AI-supported programming learning tools, we conducted a participatory design study with 15 undergraduate novice programmers and 10 instructors to gather insights on their desired help features and control preferences, as well as a follow-up survey with 172 introductory programming students. Our qualitative findings show that learners prefer help that is encouraging, incorporates visual aids, and includes peer-related insights, whereas instructors prioritize scaffolding that reflects learners' progress and reinforces best practices. Both groups favor shared control, though learners generally prefer more autonomy, while instructors lean toward greater system guidance to prevent cognitive overload. Additionally, our interviews revealed individual differences in control preferences. Based on our findings, we propose design guidelines for AI-supported programming tools, particularly regarding user-centered help features and adaptive control mechanisms. Our work contributes to the human-centered design of AI-supported learning environments by informing the development of systems that effectively balance autonomy and guidance, enhancing AI-supported educational tools for programming and beyond.

Jie Xu, Zihan Wu

Existing methods for measuring training sample influence on models only provide static, overall measurements, overlooking how sample influence changes during training. We propose Dynamic Influence Tracker (DIT), which captures the time-varying sample influence across arbitrary time windows during training. DIT offers three key insights: 1) Samples show different time-varying influence patterns, with some samples important in the early training stage while others become important later. 2) Sample influences show a weak correlation between early and late stages, demonstrating that the model undergoes distinct learning phases with shifting priorities. 3) Analyzing influence during the convergence period provides more efficient and accurate detection of corrupted samples than full-training analysis. Supported by theoretical guarantees without assuming loss convexity or model convergence, DIT significantly outperforms existing methods, achieving up to 0.99 correlation with ground truth and above 98\% accuracy in detecting corrupted samples in complex architectures.

Jie Xu, Zihan Wu, Cong Wang et al.

Time-aware data valuation enhances training efficiency and model robustness, as early detection of harmful samples could prevent months of wasted computation. However, existing methods rely on model retraining or convergence assumptions or fail to capture long-term training dynamics. We propose LiveVal, an efficient time-aware data valuation method with three key designs: 1) seamless integration with SGD training for efficient data contribution monitoring; 2) reference-based valuation with normalization for reliable benchmark establishment; and 3) adaptive reference point selection for real-time updating with optimized memory usage. We establish theoretical guarantees for LiveVal's stability and prove that its valuations are bounded and directionally aligned with optimization progress. Extensive experiments demonstrate that LiveVal provides efficient data valuation across different modalities and model scales, achieving 180 speedup over traditional methods while maintaining robust detection performance.

Aadarsh Padiyath, Xinying Hou, Amy Pang et al.

The capability of large language models (LLMs) to generate, debug, and explain code has sparked the interest of researchers and educators in undergraduate programming, with many anticipating their transformative potential in programming education. However, decisions about why and how to use LLMs in programming education may involve more than just the assessment of an LLM's technical capabilities. Using the social shaping of technology theory as a guiding framework, our study explores how students' social perceptions influence their own LLM usage. We then examine the correlation of self-reported LLM usage with students' self-efficacy and midterm performances in an undergraduate programming course. Triangulating data from an anonymous end-of-course student survey (n = 158), a mid-course self-efficacy survey (n=158), student interviews (n = 10), self-reported LLM usage on homework, and midterm performances, we discovered that students' use of LLMs was associated with their expectations for their future careers and their perceptions of peer usage. Additionally, early self-reported LLM usage in our context correlated with lower self-efficacy and lower midterm scores, while students' perceived over-reliance on LLMs, rather than their usage itself, correlated with decreased self-efficacy later in the course.