Selected work

Representative Papers

ILP-BO: Integer Linear Programming-Based Black-Box Optimization

Sep 27, 2026

This study addresses the inability of traditional black-box optimization to guarantee global optimality of candidate solutions by proposing an Integer Linear Programming (ILP)-based black-box optimization framework. The proposed method reformulates discrete-domain kernel surrogate models as ILP problems, innovatively introducing binary auxiliary variables to exactly represent nonlinear kernel functions for objective linearization. Furthermore, it incorporates a Hamming distance margin mechanism to balance exploration and exploitation. Experimental results on synthetic and discrete benchmark problems demonstrate that this approach achieves performance comparable to mainstream Bayesian optimization algorithms while delivering transparent discrete optimization with provable global optimality certificates.

0 citationsRead paper

Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment

Sep 16, 2026

研究通过多模型审计方法，探讨了开放权重大型语言模型在招聘中性别和种族偏见问题，发现特定语言风格会触发这些偏见，并提出了一种预部署审计协议。

0 citationsRead paper

Bootstrapping Niche Multilingual Code Translation via Reinforcement Learning with Execution-Based Verifiable Supervision

Aug 13, 2026

This study addresses the poor executability of code translation for low-resource languages caused by insufficient parallel supervision. We propose an execution feedback-based reinforcement learning framework that trains a reward model using execution-verified data and optimizes large language models via the GRPO algorithm to enhance cross-language code generation correctness. Additionally, we introduce Humaneval-X++, a multilingual evaluation benchmark. Experiments demonstrate that a 4B-parameter model achieves an average performance improvement of 13% on this benchmark, with gains reaching 21% for medium-resource languages. The approach successfully enables executable code translation across 600 language pairs, significantly mitigating alignment challenges in low-resource scenarios.

0 citationsRead paper

Recent publications

Latest Papers

ILP-BO: Integer Linear Programming-Based Black-Box Optimization

Sep 27, 2026

This study addresses the inability of traditional black-box optimization to guarantee global optimality of candidate solutions by proposing an Integer Linear Programming (ILP)-based black-box optimization framework. The proposed method reformulates discrete-domain kernel surrogate models as ILP problems, innovatively introducing binary auxiliary variables to exactly represent nonlinear kernel functions for objective linearization. Furthermore, it incorporates a Hamming distance margin mechanism to balance exploration and exploitation. Experimental results on synthetic and discrete benchmark problems demonstrate that this approach achieves performance comparable to mainstream Bayesian optimization algorithms while delivering transparent discrete optimization with provable global optimality certificates.

0 citationsRead paper

Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment

Sep 16, 2026

研究通过多模型审计方法，探讨了开放权重大型语言模型在招聘中性别和种族偏见问题，发现特定语言风格会触发这些偏见，并提出了一种预部署审计协议。

0 citationsRead paper

Bootstrapping Niche Multilingual Code Translation via Reinforcement Learning with Execution-Based Verifiable Supervision

Aug 13, 2026

This study addresses the poor executability of code translation for low-resource languages caused by insufficient parallel supervision. We propose an execution feedback-based reinforcement learning framework that trains a reward model using execution-verified data and optimizes large language models via the GRPO algorithm to enhance cross-language code generation correctness. Additionally, we introduce Humaneval-X++, a multilingual evaluation benchmark. Experiments demonstrate that a 4B-parameter model achieves an average performance improvement of 13% on this benchmark, with gains reaching 21% for medium-resource languages. The approach successfully enables executable code translation across 600 language pairs, significantly mitigating alignment challenges in low-resource scenarios.

0 citationsRead paper