arXiv subjects

Jiajun Sun

Publications and source records attributed to Jiajun Sun.

At least 37 records · Page 2Linked to original sources

arXiv2026

CMDAR: A Chinese Multi-scene Dynamic Audio Reasoning Benchmark with Diverse Challenges

The ability to reason from audio, including speech, environmental sounds, and music, is essential for AI agents to interact effectively in real-world scenarios. Existing benchmarks mainly focus on static or single-scene settings and English audio data and do not fully capture scenarios where multiple speakers, unfolding events, and heterogeneous audio sources interact. To address these challenges, we introduce CMDAR, a Chinese benchmark for evaluating models on complex, multi-scene, and dynamically evolving audio reasoning tasks. CMDAR comprises 3,000 carefully curated question-answer pairs linked to diverse audio clips, covering five categories of complex reasoning and spanning three question types. We benchmark 26 state-of-the-art audio language models on CMDAR and observe that they exhibit limitations in complex reasoning tasks. In CMDAR-main, Qwen2.5-Omni achieves 76.67% accuracy, whereas GPT-4o Audio reaches 68.47%. However, GPT-4o Audio substantially outperforms Qwen2.5-Omni on the more challenging multiple-choice with multiple audios and open-ended tasks. And we provide detail analysis corresponding suggestions for the future development of large audio language models.

cs.SD↗

arXiv2025

A General Stability Approach to False Discovery Rate Control

Stability and reproducibility are essential considerations in various applications of statistical methods. False Discovery Rate (FDR) control methods are able to control false signals in scientific discoveries. However, many FDR control methods, such as Model-X knockoff and data-splitting approaches, yield unstable results due to the inherent randomness of the algorithms. To enhance the stability and reproducibility of statistical outcomes, we propose a general stability approach for FDR control in feature selection and multiple testing problems, named FDR Stabilizer. Taking feature selection as an example, our method first aggregates feature importance statistics obtained by multiple runs of the base FDR control procedure into a consensus ranking. Then, we construct a stabilized relaxed e-value for each feature and apply the e-BH procedure to these stabilized e-values to obtain the final selection set. We theoretically derive the finite-sample bounds for the FDR and the power of our method, and show that our method asymptotically controls the FDR without power loss. Moreover, we establish the stability of the proposed method, showing that the stabilized selection set converges to a deterministic limit as the number of repetitions increases. Extensive numerical experiments and applications to real datasets demonstrate that the proposed method generally outperforms existing alternatives.

stat.ME↗

arXiv2025

The Role of Entropy in Visual Grounding: Analysis and Optimization

Recent advances in fine-tuning multimodal large language models (MLLMs) using reinforcement learning have achieved remarkable progress, particularly with the introduction of various entropy control techniques. However, the role and characteristics of entropy in perception-oriented tasks like visual grounding, as well as effective strategies for controlling it, remain largely unexplored. To address this issue, we focus on the visual grounding task and analyze the role and characteristics of entropy in comparison to reasoning tasks. Building on these findings, we introduce ECVGPO (Entropy Control Visual Grounding Policy Optimization), an interpretable algorithm designed for effective entropy regulation. Through entropy control, the trade-off between exploration and exploitation is better balanced. Experiments show that ECVGPO achieves broad improvements across various benchmarks and models.

cs.CV↗

arXiv2025

From Scores to Preferences: Redefining MOS Benchmarking for Speech Quality Reward Modeling

Assessing the perceptual quality of synthetic speech is crucial for guiding the development and refinement of speech generation models. However, it has traditionally relied on human subjective ratings such as the Mean Opinion Score (MOS), which depend on manual annotations and often suffer from inconsistent rating standards and poor reproducibility. To address these limitations, we introduce MOS-RMBench, a unified benchmark that reformulates diverse MOS datasets into a preference-comparison setting, enabling rigorous evaluation across different datasets. Building on MOS-RMBench, we systematically construct and evaluate three paradigms for reward modeling: scalar reward models, semi-scalar reward models, and generative reward models (GRMs). Our experiments reveal three key findings: (1) scalar models achieve the strongest overall performance, consistently exceeding 74% accuracy; (2) most models perform considerably worse on synthetic speech than on human speech; and (3) all models struggle on pairs with very small MOS differences. To improve performance on these challenging pairs, we propose a MOS-aware GRM that incorporates an MOS-difference-based reward function, enabling the model to adaptively scale rewards according to the difficulty of each sample pair. Experimental results show that the MOS-aware GRM significantly improves fine-grained quality discrimination and narrows the gap with scalar models on the most challenging cases. We hope this work will establish both a benchmark and a methodological framework to foster more rigorous and scalable research in automatic speech quality assessment.

cs.SD↗

arXiv2025

Controllable Skin Synthesis via Lesion-Focused Vector Autoregression Model

Skin images from real-world clinical practice are often limited, resulting in a shortage of training data for deep-learning models. While many studies have explored skin image synthesis, existing methods often generate low-quality images and lack control over the lesion's location and type. To address these limitations, we present LF-VAR, a model leveraging quantified lesion measurement scores and lesion type labels to guide the clinically relevant and controllable synthesis of skin images. It enables controlled skin synthesis with specific lesion characteristics based on language prompts. We train a multiscale lesion-focused Vector Quantised Variational Auto-Encoder (VQVAE) to encode images into discrete latent representations for structured tokenization. Then, a Visual AutoRegressive (VAR) Transformer trained on tokenized representations facilitates image synthesis. Lesion measurement from the lesion region and types as conditional embeddings are integrated to enhance synthesis fidelity. Our method achieves the best overall FID score (average 0.74) among seven lesion types, improving upon the previous state-of-the-art (SOTA) by 6.3%. The study highlights our controllable skin synthesis model's effectiveness in generating high-fidelity, clinically relevant synthetic skin images. Our framework code is available at https://github.com/echosun1996/LF-VAR.

cs.CV↗

arXiv2025

Empirical Evaluation of Large Language Models in Automated Program Repair

The increasing prevalence of software bugs has made automated program repair (APR) a key research focus. Large language models (LLMs) offer new opportunities for APR, but existing studies mostly rely on smaller, earlier-generation models and Java benchmarks. The repair capabilities of modern, large-scale LLMs across diverse languages and scenarios remain underexplored. To address this, we conduct a comprehensive empirical study of four open-source LLMs, CodeLlama, LLaMA, StarCoder, and DeepSeek-Coder, spanning 7B to 33B parameters, diverse architectures, and purposes. We evaluate them across two bug scenarios (enterprise-grades and algorithmic), three languages (Java, C/C++, Python), and four prompting strategies, analyzing over 600K generated patches on six benchmarks. Key findings include: (1) model specialization (e.g., CodeLlama) can outperform larger general-purpose models (e.g., LLaMA); (2) repair performance does not scale linearly with model size; (3) correct patches often appear early in generation; and (4) prompts significantly affect results. These insights offer practical guidance for designing effective and efficient LLM-based APR systems.

cs.SE↗

arXiv2025

Mitigating Object Hallucinations in MLLMs via Multi-Frequency Perturbations

Recently, multimodal large language models (MLLMs) have demonstrated remarkable performance in visual-language tasks. However, the authenticity of the responses generated by MLLMs is often compromised by object hallucinations. We identify that a key cause of these hallucinations is the model's over-susceptibility to specific image frequency features in detecting objects. In this paper, we introduce Multi-Frequency Perturbations (MFP), a simple, cost-effective, and pluggable method that leverages both low-frequency and high-frequency features of images to perturb visual feature representations and explicitly suppress redundant frequency-domain features during inference, thereby mitigating hallucinations. Experimental results demonstrate that our method significantly mitigates object hallucinations across various model architectures. Furthermore, as a training-time method, MFP can be combined with inference-time methods to achieve state-of-the-art performance on the CHAIR benchmark.

cs.CV↗

arXiv2025

Evaluating the Generalizability of LLMs in Automated Program Repair

LLM-based automated program repair methods have attracted significant attention for their state-of-the-art performance. However, they were primarily evaluated on a few well known datasets like Defects4J, raising questions about their effectiveness on new datasets. In this study, we evaluate 11 top-performing LLMs on DEFECTS4J-TRANS, a new dataset derived from transforming Defects4J while maintaining the original semantics. Results from experiments on both Defects4J and DEFECTS4J-TRANS show that all studied LLMs have limited generalizability in APR tasks, with the average number of correct and plausible patches decreasing by 49.48% and 42.90%, respectively, on DEFECTS4J-TRANS. Further investigation into incorporating additional repair-relevant information in repair prompts reveals that, although this information significantly enhances the LLMs' capabilities (increasing the number of correct and plausible patches by up to 136.67% and 121.82%, respectively), performance still falls short of their original results. This indicates that prompt engineering alone is insufficient to substantially enhance LLMs' repair capabilities. Based on our study, we also offer several recommendations for future research.

cs.SE↗

arXiv2024

Hybrid Automated Program Repair by Combining Large Language Models and Program Analysis

Automated Program Repair (APR) has garnered significant attention due to its potential to streamline the bug repair process for human developers. Recently, LLM-based APR methods have shown promise in repairing real-world bugs. However, existing APR methods often utilize patches generated by LLMs without further optimization, resulting in reduced effectiveness due to the lack of program-specific knowledge. Furthermore, the evaluations of these APR methods have typically been conducted under the assumption of perfect fault localization, which may not accurately reflect their real-world effectiveness. To address these limitations, this paper introduces an innovative APR approach called GIANTREPAIR. Our approach leverages the insight that LLM-generated patches, although not necessarily correct, offer valuable guidance for the patch generation process. Based on this insight, GIANTREPAIR first constructs patch skeletons from LLM-generated patches to confine the patch space, and then generates high-quality patches tailored to specific programs through context-aware patch generation by instantiating the skeletons. To evaluate the performance of our approach, we conduct two large-scale experiments. The results demonstrate that GIANTREPAIR not only effectively repairs more bugs (an average of 27.78% on Defects4J v1.2 and 23.40% on Defects4J v2.0) than using LLM-generated patches directly, but also outperforms state-of-the-art APR methods by repairing at least 42 and 7 more bugs under perfect and automated fault localization scenarios, respectively.

cs.SE↗

arXiv2022

DATA: Domain-Aware and Task-Aware Self-supervised Learning

The paradigm of training models on massive data without label through self-supervised learning (SSL) and finetuning on many downstream tasks has become a trend recently. However, due to the high training costs and the unconsciousness of downstream usages, most self-supervised learning methods lack the capability to correspond to the diversities of downstream scenarios, as there are various data domains, different vision tasks and latency constraints on models. Neural architecture search (NAS) is one universally acknowledged fashion to conquer the issues above, but applying NAS on SSL seems impossible as there is no label or metric provided for judging model selection. In this paper, we present DATA, a simple yet effective NAS approach specialized for SSL that provides Domain-Aware and Task-Aware pre-training. Specifically, we (i) train a supernet which could be deemed as a set of millions of networks covering a wide range of model scales without any label, (ii) propose a flexible searching mechanism compatible with SSL that enables finding networks of different computation costs, for various downstream vision tasks and data domains without explicit metric provided. Instantiated With MoCo v2, our method achieves promising results across a wide range of computation costs on downstream tasks, including image classification, object detection and semantic segmentation. DATA is orthogonal to most existing SSL methods and endows them the ability of customization on downstream needs. Extensive experiments on other SSL methods demonstrate the generalizability of the proposed method. Code is released at https://github.com/GAIA-vision/GAIA-ssl

cs.CV↗

arXiv2015

Long-Term Profit-Maximizing Incentive for Crowd Sensing in Mobile Social Networks

Crowd sensing is a new paradigm that leverages pervasive sensor-equipped mobile devices to provide sensing services like forensic analysis, documenting public spaces, and collaboratively constructing statistical models. Extensive user participation is indispensable for achieving good service quality. Nowadays, most of existing mechanisms focus on guaranteeing good service quality based on instantaneous extensive user participation for crowd sensing applications. Little attention has been dedicated to maximizing long-term service quality for crowd sensing applications due to their asymmetric interests, preferences, selfish behaviors, etc. To fill these gaps, in this paper, we derive the closed expression of the marginal sensing data quality based on the monopoly aggregation in economics. Furthermore, we design marginalquality based incentive mechanisms for long-term crowd sensing applications, not only to enhance extensive user participation by maximizing the expected total profits of mobile users, but also to stimulate mobile users to produce high-quality contents by applying the marginal quality. Finally, simulation results show that our mechanisms outperform the existing solutions.

cs.SI↗

arXiv2015

An Incentive Mechanism for Periodical Mobile Crowdsensing from a Frugality Perspective

Mobile crowdsensing (MCS) has been intensively explored recently due to its flexible and pervasive sensing ability. Although many incentive mechanisms have been built to attract extensive user participation, Most of these mechanisms focus only on independent task scenarios, where the sensing tasks are independent of each other. On the contrary, we focus on a periodical task scenario, where each user participates in the same type of sensing tasks periodically. In this paper, we consider the long-term user participation incentive in a general periodical MCS system from a frugality payment perspective. We explore the issue under both semi-online (the intra-period interactive process is synchronous while the inter-period interactive process is sequential and asynchronous during each period) and online user arrival models (the previous two interactive processes are sequential and asynchronous). In particular, we first propose a semi-online frugal incentive mechanism by introducing a Lyapunov method. Moreover, we also extend it to an online frugal incentive mechanism, which satisfies the constant frugality. Besides, the two mechanisms can also satisfy computational efficiency, asymptotical optimality, individual rationality and truthfulness. Through extensive simulations, we evaluate the performance and validate the theoretical properties of our online mechanisms.

cs.GT↗

arXiv2014

Service-Constraint Based Truthful Incentive Mechanisms for Crowd Sensing

Crowd sensing is a new paradigm which leverages the pervasive smartphones to efficiently collect and upload sensing data, enabling numerous novel applications. To achieve good service quality for a crowd sensing application, incentive mechanisms are necessary for attracting more user participation. Most of existing mechanisms apply only for the budget-constraint scenario where the platform (the crowd sensing organizer) has a budget limit. On the contrary, we focus on a different scenario where the platform has a service limit. Based on the offline and online auction model, we consider a general problem: users submit their private profiles to the platform, and the platform aims at selecting a subset of users before a specified deadline for minimizing the total payment while a specific service can be completed. Specially, we design offline and online service-constraint incentive mechanisms for the case where the value function of selected users is monotone submodular. The mechanisms are individual rationality, task feasibility, computational efficiency, truthfulness, consumer sovereignty, constant frugality, and also performs well in practice. Finally, we use extensive simulations to demonstrate the theoretical properties of our mechanisms.

cs.NI↗

arXiv2014

Privacy-Preserving Verifiable Incentive Mechanism for Crowdsourcing Market Applications

Recently, a novel class of incentive mechanisms is proposed to attract extensive users to truthfully participate in crowd sensing applications with a given budget constraint. The class mechanisms also bring good service quality for the requesters in crowd sensing applications. Although it is so important, there still exists many verification and privacy challenges, including users' bids and subtask information privacy and identification privacy, winners' set privacy of the platform, and the security of the payment outcomes. In this paper, we present a privacy-preserving verifiable incentive mechanism for crowd sensing applications with the budget constraint, not only to explore how to protect the privacies of users and the platform, but also to make the verifiable payment correct between the platform and users for crowd sensing applications. Results indicate that our privacy-preserving verifiable incentive mechanism achieves the same results as the generic one without privacy preservation.

cs.GT↗

arXiv2014

Incentive Mechanisms for Mobile Crowd Sensing: Current States and Challenges of Work

Mobile crowd sensing (MCS) is a new paradigm which leverages the ubiquity of sensor-equipped mobile devices such as smartphones, music players, and in-vehicle sensors at the edge of the Internet, to collect data. The new paradigm will fuel the evolution of the Internet of Things to three changes as follows: First, the terminal devices at the edge of the Internet change from PCs to mobile phones. Second, the interactive mode extends from the virtual space to the real physical world. Thirdly, the forwarding manner of sensing data are undergoing the transition from the priori to the opportunistic. To better meet the demands of MCS applications at a societal scale, incentive mechanisms are indispensable. In this paper, we will first overview three categories of MCS applications, and then propose a new architecture for MCS applications. Based on the architecture, we discuss various research challenges about incentive mechanism designs for MCS applications, followed by potential future work discussions. Finally, we present potential future works.

cs.NI↗

arXiv2014

Collection-behavior based Multi-parameter Posted Pricing Mechanism for Crowd Sensing

Crowd sensing is a new paradigm which leverages a large number of sensor-equipped mobile phones to collect sensing data. To attract more participants to provide good quality, bidding mechanisms that solicit the Vickrey-Clarke-Groves (VCG) mechanism and its variants are natural fits for crowd sensing applications in mobile social networks. However, in practical continuous crowd sensing applications, where bids cannot be solicited and only posted pricing mechanisms can be implemented. Obviously, these mechanisms for continuous crowd sensing are not applicable. To tackle the issue, we propose a collection-behavior based multi-parameter posted pricing mechanism, not only to consider extensive user participating and sensing data submission quality under given budget constraints by applying all-pay auctions and posted pricing mechanisms, but also to maximize the coverage utilities by applying crowd aversion. Simulation results indicate that incentive mechanisms in our proposed framework outperform the best existing solution.

cs.GT↗

arXiv2014

General Privacy-Preserving Verifiable Incentive Mechanism for Crowdsourcing Markets

In crowdsourcing markets, there are two different type jobs, i.e. homogeneous jobs and heterogeneous jobs, which need to be allocated to workers. Incentive mechanisms are essential to attract extensive user participating for achieving good service quality, especially under a given budget constraint condition. To this end, recently, Singer et al. propose a novel class of auction mechanisms for determining near-optimal prices of tasks for crowdsourcing markets constrained by the given budget. Their mechanisms are very useful to motivate extensive user to truthfully participate in crowdsourcing markets. Although they are so important, there still exist many security and privacy challenges in real-life environments. In this paper, we present a general privacy-preserving verifiable incentive mechanism for crowdsourcing markets with the budget constraint, not only to exploit how to protect the bids and assignments' privacy, and the chosen winners' privacy in crowdsourcing markets with homogeneous jobs and heterogeneous jobs and identity privacy from users, but also to make the verifiable payment between the platform and users for crowdsourcing applications. Results show that our general privacy-preserving verifiable incentive mechanisms achieve the same results as the generic one without privacy preservation.

cs.GT↗

arXiv2014

Behavior-Based online Incentive Mechanism for Crowd Sensing with Budget Constraints

Crowd sensing is a new paradigm which leverages the ubiquity of sensor-equipped mobile devices to collect data. To achieve good quality for crowd sensing, incentive mechanisms are indispensable to attract more participants. Most of existing mechanisms focus on the expected utility prior to sensing, ignoring the risk of low quality solution and privacy leakage. Traditional incentive mechanisms such as the Vickrey-Clarke-Groves (VCG) mechanism and its variants are not applicable here. In this paper, to address these challenges, we propose a behavior based incentive mechanism for crowd sensing applications with budget constraints by applying sequential all-pay auctions in mobile social networks (MSNs), not only to consider the effects of extensive user participation, but also to maximize high quality of the context based sensing content submission for crowd sensing platform under the budget constraints, where users arrive in a sequential order. Through an extensive simulation, results indicate that incentive mechanisms in our proposed framework outperform the best existing solution.

cs.GT↗