Tianlin Li

Dr. Tianlin Li


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 October 2025 in arXiv
Verified
Preprint
Open access
PDF file
KWKaixin Wang
Kaixin Wang
TLTianlin Li
Tianlin Li
XZXiaoyu Zhang
Xiaoyu Zhang
ALAishan Liu
Aishan Liu
XLXianglong Liu
Xianglong Liu
Bin ShiBin Shi
Bin Shi
Code Large Language Models (CodeLLMs) are increasingly used in code generation tasks across a wide range of applications. However, their performance is often inconsistent across different programming languages (PLs), with low-resource PLs suffering the most due to limited training data. In this paper, we present CodeChemist, a novel and efficient framework for test-time scaling that enables functional knowledge transfer from high-resource to low-resource PLs using generated test cases. CodeChemist first generates and executes code in high-resource PLs to create test cases that encapsulate functional knowledge. It then uses multi-temperature hedged sampling to generate code snippets in the low-resource PL and selects the best one based on the pass rate of the test cases. Our extensive experiments show that CodeChemist outperforms existing test-time scaling approaches, boosting the performance of code generation for low-resource PLs without requiring any model retraining.

22 October 2025 in arXiv
Verified
Preprint
Open access
PDF file
QXQian Xiong
Qian Xiong
BYBo Yang
Bo Yang
Weisong SunWeisong Sun
Weisong Sun
YZYiran Zhang
Yiran Zhang
TLTianlin Li
Tianlin Li
ZJZhi Jin
Zhi Jin
Automated code generation driven by Large Lan- guage Models (LLMs) has enhanced development efficiency, yet generating complex application-level software code remains challenging. Multi-agent frameworks show potential, but existing methods perform inadequately in large-scale application-level software code generation, failing to ensure reasonable orga- nizational structures of project code and making it difficult to maintain the code generation process. To address this, this paper envisions a Knowledge-Guided Application-Level Code Generation framework named KGACG, which aims to trans- form software requirements specification and architectural design document into executable code through a collaborative closed- loop of the Code Organization & Planning Agent (COPA), Coding Agent (CA), and Testing Agent (TA), combined with a feedback mechanism. We demonstrate the collaborative process of the agents in KGACG in a Java Tank Battle game case study while facing challenges. KGACG is dedicated to advancing the automation of application-level software development.

9 November 2026 in Electronics
Verified
Journal Article
Open access
PDF file
LYLi Yangchun
Li Yangchun
TLTianlin Li
Tianlin Li
MCMingfei Cheng
Mingfei Cheng
YLYan Li
Yan Li
DYDanfeng Yan
Danfeng Yan
Large vision–language models (VLMs) can infer geographic information from visual content even when explicit metadata are absent. We evaluate this capability as a privacy-relevant proxy: country, sub-national, city, and coordinate accuracy measure potential location exposure, not realized privacy harm. VideoGeoEval is a 1000-sequence diagnostic benchmark constructed from BDD100K driving videos and CityGuessr multi-view samples. We query four proprietary API models with one- and eight-frame inputs under Base, text-masked, blurred, and geographic-text-spoofed conditions. A containment-aware scorer enforces Country ≥ State ≥ City, and coordinate results satisfy the nested 50/200 km thresholds. Across three runs, eight-frame input improves city accuracy by 4.5–8.0 percentage points on CityGuessr and 5.5–7.1 points on BDD100K, with paired bootstrap confidence intervals excluding zero. Prior-normalized results indicate that the BDD100K gains cannot be fully explained by dominant-label priors. Spoofed overlays substantially reduce accuracy, but are treated only as controlled cross-modal stress tests. The benchmark therefore characterizes potential location exposure under controlled queries while separating measured accuracy from downstream privacy consequences.

17 May 2025 in arXiv
Verified
Preprint
Open access
PDF file
QZQi Zhou
Qi Zhou
JZJie Zhang
Jie Zhang
DWDongxia Wang
Dongxia Wang
QLQiang Liu
Qiang Liu
TLTianlin Li
Tianlin Li
Qing GuoQing Guo
Qing Guo
Human preference plays a crucial role in the refinement of large language models (LLMs). However, collecting human preference feedback is costly and most existing datasets neglect the correlation between personalization and preferences. To address this issue, we introduce Fair-PP, a synthetic dataset of personalized preferences targeting social equity, derived from real-world social survey data, which includes 28 social groups, 98 equity topics, and 5 personal preference dimensions. Leveraging GPT-4o-mini, we engage in role-playing based on seven representative persona portrayals guided by existing social survey data, yielding a total of 238,623 preference records. Through Fair-PP, we also contribute (i) An automated framework for generating preference data, along with a more fine-grained dataset of personalized preferences; (ii) analysis of the positioning of the existing mainstream LLMs across five major global regions within the personalized preference space; and (iii) a sample reweighting method for personalized preference alignment, enabling alignment with a target persona while maximizing the divergence from other personas. Empirical experiments show our method outperforms the baselines.

9 August 2025 in arXiv
Verified
Preprint
Open access
PDF file
CLChangqing Li
Changqing Li
TLTianlin Li
Tianlin Li
Xiaohan ZhangXiaohan Zhang
Xiaohan Zhang
Aishan LiuAishan Liu
Aishan Liu
LPLi Pan
Li Pan
Large Language Models (LLMs) face persistent and evolving trustworthiness issues, motivating developers to seek automated and flexible repair methods that enable convenient deployment across diverse scenarios. Existing repair methods like supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) are costly and slow, while prompt engineering lacks robustness and scalability. Representation engineering, which steers model behavior by injecting targeted concept vectors during inference, offers a lightweight, training-free alternative. However, current approaches depend on manually crafted samples and fixed steering strategies, limiting automation and adaptability. To overcome these challenges, we propose MASteer, the first end-to-end framework for trustworthiness repair in LLMs based on representation engineering. MASteer integrates two core components: AutoTester, a multi-agent system that generates diverse, high-quality steer samples tailored to developer needs; and AutoRepairer, which constructs adaptive steering strategies with anchor vectors for automated, context-aware strategy selection during inference. Experiments on standard and customized trustworthiness tasks show MASteer consistently outperforms baselines, improving metrics by 15.36% on LLaMA-3.1-8B-Chat and 4.21% on Qwen-3-8B-Chat, while maintaining general model capabilities. MASteer demonstrates strong robustness, generalization, and practical value for scalable, efficient trustworthiness repair.

6 October 2025 in arXiv
Verified
Preprint
Open access
PDF file
JZJingyao Zhang
Jingyao Zhang
TLTianlin Li
Tianlin Li
XZXiaoyu Zhang
Xiaoyu Zhang
QHQiang Hu
Qiang Hu
Bin ShiBin Shi
Bin Shi
Autoregressive Large Language Models (AR-LLMs) are widely used in software engineering (SE) but face limitations in processing code structure information and suffer from high inference latency. Diffusion LLMs (DLLMs) offer a promising alternative with global bidirectional encoding and decoupled generation steps. This work presents the first comprehensive evaluation of DLLMs across the software development lifecycle, including code generation, defect detection, and program repair. On a large-scale benchmark of 52,937 tasks, 7Bparameter DLLMs outperform AR-LLMs with a 30% average accuracy improvement achieving a 113% gain on cross-file repair, while maintaining superior efficiency and reduced latency. Our results establish DLLMs as a superior paradigm for SE tasks.