Selected work

Representative Papers

An Empirical Analysis of Compatibility Issues for Industrial Mobile Games (Practical Experience Report)

Oct 01, 2022IEEE International Symposium on Software Reliability Engineering

Mobile games face severe compatibility issues amid extreme device and OS version fragmentation, yet existing research lacks systematic empirical analysis. This study conducts a large-scale industrial empirical investigation, constructing the first production-grade mobile game defect dataset—comprising 7 million lines of code and 20,000 commits. Through rigorous manual annotation, root-cause categorization, and repair-strategy clustering, we systematically identify six primary compatibility root-cause categories and their corresponding repair patterns—the first such classification in the literature. We further propose reusable diagnostic guidelines and proactive prevention recommendations, demonstrably improving incident response efficiency. This work fills a critical gap in empirical software engineering research on industrial mobile game compatibility, providing both theoretical foundations and practical paradigms for testing, diagnosis, and remediation.

2 citationsRead paper

Prompt Candidates, then Distill: A Teacher-Student Framework for LLM-driven Data Annotation

Jun 04, 2025

To address the problem that large language models (LLMs) often mislabel ambiguous or difficult instances due to inherent uncertainty—thereby degrading downstream task performance—this paper proposes a novel “candidate labeling” paradigm. Instead of generating a single deterministic label, the LLM outputs all plausible candidate labels; a lightweight small language model (SLM) then performs uncertainty-aware supervised distillation to produce a high-quality single-label prediction. Inspired by human cognitive strategies for avoiding ambiguity, this approach is theoretically shown to achieve superior statistical consistency and tighter generalization bounds compared to conventional single-label annotation. Implemented within a teacher–student framework (CanDist) and enhanced with uncertainty-driven prompting, our method improves annotation quality across six text classification benchmarks, boosting downstream model accuracy by 2.1–4.7 percentage points on average. The code is publicly available.

1 citationsRead paper

AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems

Sep 04, 2026

为解决工业推荐系统中从研究到部署的自动化问题，AutoLR通过多专家委员会、确定性探索-利用选择器和分层知识系统相结合的方法实现全流程自动化。

0 citationsRead paper

ES3D: Embedding Semantics into 3D Space for Component-Aware Editing

Aug 16, 2026

This study addresses the limitations of fine-grained semantic understanding and precise component manipulation in existing 3D editing approaches by proposing a novel 3D semantic embedding framework. Leveraging voxelized semantic projection and feature similarity retrieval, the method achieves component-level perception without manual masking. Furthermore, it integrates multi-reference image guidance, text-conditioned control, and generative inpainting mechanisms to enable refined editing. This approach effectively overcomes bottlenecks in specific component retrieval and modification while facilitating robust multimodal conditional control. Consequently, the proposed framework significantly enhances both geometric consistency and semantic coherence in editing outcomes, establishing a new paradigm for the fine-grained generation of 3D assets.

0 citationsRead paper

Confucius4-TTS: Transcript-Free Cross-Lingual Zero-Shot TTS with a Learnable Speaker Encoder

Aug 12, 2026

This work addresses the limitation of existing zero-shot text-to-speech (TTS) systems in cross-lingual voice cloning, which typically rely on transcriptions of reference audio and thus struggle in real-world scenarios where such transcripts are unavailable. To overcome this, the authors propose a transcription-free multilingual zero-shot TTS system featuring a two-stage architecture: a large language model–based text-to-semantic generation module followed by a conditional flow matching–based semantic-to-acoustic synthesis module. A learnable speaker encoder is introduced to extract prosodic and timbral characteristics directly from self-supervised speech representations, enabling both intra- and cross-lingual voice cloning across 14 languages. Experimental results on the CV3-Eval benchmark show an average word error rate of 3.73%, and human evaluations indicate superior overall performance compared to current open-source and commercial systems.

0 citationsRead paper

Recent publications

Latest Papers

AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems

Sep 04, 2026

为解决工业推荐系统中从研究到部署的自动化问题，AutoLR通过多专家委员会、确定性探索-利用选择器和分层知识系统相结合的方法实现全流程自动化。

0 citationsRead paper

ES3D: Embedding Semantics into 3D Space for Component-Aware Editing

Aug 16, 2026

This study addresses the limitations of fine-grained semantic understanding and precise component manipulation in existing 3D editing approaches by proposing a novel 3D semantic embedding framework. Leveraging voxelized semantic projection and feature similarity retrieval, the method achieves component-level perception without manual masking. Furthermore, it integrates multi-reference image guidance, text-conditioned control, and generative inpainting mechanisms to enable refined editing. This approach effectively overcomes bottlenecks in specific component retrieval and modification while facilitating robust multimodal conditional control. Consequently, the proposed framework significantly enhances both geometric consistency and semantic coherence in editing outcomes, establishing a new paradigm for the fine-grained generation of 3D assets.

0 citationsRead paper

Confucius4-TTS: Transcript-Free Cross-Lingual Zero-Shot TTS with a Learnable Speaker Encoder

Aug 12, 2026

This work addresses the limitation of existing zero-shot text-to-speech (TTS) systems in cross-lingual voice cloning, which typically rely on transcriptions of reference audio and thus struggle in real-world scenarios where such transcripts are unavailable. To overcome this, the authors propose a transcription-free multilingual zero-shot TTS system featuring a two-stage architecture: a large language model–based text-to-semantic generation module followed by a conditional flow matching–based semantic-to-acoustic synthesis module. A learnable speaker encoder is introduced to extract prosodic and timbral characteristics directly from self-supervised speech representations, enabling both intra- and cross-lingual voice cloning across 14 languages. Experimental results on the CV3-Eval benchmark show an average word error rate of 3.73%, and human evaluations indicate superior overall performance compared to current open-source and commercial systems.

0 citationsRead paper

ASR-Roundtrip Evaluation Can Mask Context- and Convention-Dependent Reading Errors in Chinese News TTS

Aug 11, 2026

This study addresses a critical limitation of ASR-roundtrip evaluation in Chinese news text-to-speech (TTS) systems, where context- or domain-convention-dependent reading errors are often masked, leading to perceptible errors being misjudged as correct. The work presents the first systematic analysis of this blind spot and introduces a segment-isolation diagnostic technique that combines human annotation with multi-system comparisons—employing MiMo and CosyVoice TTS systems alongside Qwen3-ASR and Paraformer ASR—to re-expose previously concealed errors. Among 110 high-risk cases, 97 were confirmed as actual errors; Qwen3-ASR superficially recovered 40 of them, whereas Paraformer recovered only 2. These findings demonstrate that ASR-roundtrip evaluation alone is insufficient for assessing Chinese news TTS quality and underscore the necessity of context-sensitive error detection.

0 citationsRead paper

Is Personalized Modality Weighting Actually Personalized? A Controlled Audit of Per-User Weighting Claims in Multimodal Recommenders

Aug 06, 2026

This work investigates whether user-specific modality weighting mechanisms widely adopted in multimodal recommendation systems genuinely capture individual user preferences. To this end, the authors propose an auditing framework comprising two metrics—real-GM and real-shuf—that evaluate personalization efficacy by comparing six personalized weighting methods against global weights and shuffled user-weight assignments, all under a unified collaborative filtering backbone. Experimental results across three short-video and one cross-domain e-commerce dataset reveal that performance gains from most methods stem primarily from increased model capacity rather than authentic user signals, with gating mechanisms often inducing spurious personalization due to their reliance on shared embeddings. Notably, global modality weights already achieve nearly all attainable gains, while personalized weighting shows no consistent improvement; the proposed audit framework effectively identifies architectures that truly encode user-specific patterns.

0 citationsRead paper