Selected work

Representative Papers

Hallucination Reduction for LLM-Based Audio Understanding via Multimodal Direct Preference Optimization

Oct 02, 2026

This study addresses the tendency of large audio-language models to hallucinate and over-rely on textual priors when processing audio-visual inputs. To mitigate this, we propose a multimodal Direct Preference Optimization (DPO) method tailored for the audio domain. Using Qwen2-Audio as the backbone, our approach systematically evaluates acoustic perturbation strategies—such as temporal reversal and frequency masking—to identify optimal configurations for constructing raw-versus-distorted audio preference pairs, thereby compelling the model to generate content grounded in authentic acoustic features. Experimental results demonstrate that the proposed method improves accuracy by 14.0% on the DCASE dataset and enhances existence verification performance on Audio Hallucination by 27.4%. These findings indicate that our approach significantly strengthens the model's temporal reasoning capabilities while effectively suppressing hallucination issues.

0 citationsRead paper

Enhancing Large Language Models with Retrieval Augmented Generation for Software Testing and Inspection Automation

Apr 16, 2026

This work addresses the challenge of unreliable outputs from large language models in software test case generation and code review due to hallucination. It presents the first systematic integration of Retrieval-Augmented Generation (RAG) into software verification and validation activities. By constructing a RAG pipeline that dynamically incorporates contextual information from external knowledge bases, the approach effectively mitigates model hallucinations and enhances the accuracy and reliability of generated content. Experimental results demonstrate that this method significantly improves the quality of both test case generation and code review, reducing manual effort while strengthening the overall effectiveness of software verification processes.

0 citationsRead paper

Prismatoid Band-Unfolding Revisited

Mar 10, 2026

This study investigates whether prismatoids admit non-overlapping edge unfoldings, with a focus on the band unfolding problem for nested prismatoids. By integrating geometric analysis and combinatorial topology, and employing a hybrid strategy that combines petal unfolding with band unfolding, the work establishes—for the first time—the necessary and sufficient conditions under which the band unfolding of a nested prismatoid is overlap-free. It further demonstrates that all known counterexamples to such unfoldings are, in essence, the only possible ones. Although the class of unfoldable shapes is not expanded, this research provides crucial theoretical tools that deepen the understanding of unfolding mechanisms and lay a foundation for addressing the more general, non-nested case.

0 citationsRead paper

Recent publications

Latest Papers

Hallucination Reduction for LLM-Based Audio Understanding via Multimodal Direct Preference Optimization

Oct 02, 2026

This study addresses the tendency of large audio-language models to hallucinate and over-rely on textual priors when processing audio-visual inputs. To mitigate this, we propose a multimodal Direct Preference Optimization (DPO) method tailored for the audio domain. Using Qwen2-Audio as the backbone, our approach systematically evaluates acoustic perturbation strategies—such as temporal reversal and frequency masking—to identify optimal configurations for constructing raw-versus-distorted audio preference pairs, thereby compelling the model to generate content grounded in authentic acoustic features. Experimental results demonstrate that the proposed method improves accuracy by 14.0% on the DCASE dataset and enhances existence verification performance on Audio Hallucination by 27.4%. These findings indicate that our approach significantly strengthens the model's temporal reasoning capabilities while effectively suppressing hallucination issues.

0 citationsRead paper

Enhancing Large Language Models with Retrieval Augmented Generation for Software Testing and Inspection Automation

Apr 16, 2026

This work addresses the challenge of unreliable outputs from large language models in software test case generation and code review due to hallucination. It presents the first systematic integration of Retrieval-Augmented Generation (RAG) into software verification and validation activities. By constructing a RAG pipeline that dynamically incorporates contextual information from external knowledge bases, the approach effectively mitigates model hallucinations and enhances the accuracy and reliability of generated content. Experimental results demonstrate that this method significantly improves the quality of both test case generation and code review, reducing manual effort while strengthening the overall effectiveness of software verification processes.

0 citationsRead paper

Prismatoid Band-Unfolding Revisited

Mar 10, 2026

This study investigates whether prismatoids admit non-overlapping edge unfoldings, with a focus on the band unfolding problem for nested prismatoids. By integrating geometric analysis and combinatorial topology, and employing a hybrid strategy that combines petal unfolding with band unfolding, the work establishes—for the first time—the necessary and sufficient conditions under which the band unfolding of a nested prismatoid is overlap-free. It further demonstrates that all known counterexamples to such unfoldings are, in essence, the only possible ones. Although the class of unfoldable shapes is not expanded, this research provides crucial theoretical tools that deepen the understanding of unfolding mechanisms and lay a foundation for addressing the more general, non-nested case.

0 citationsRead paper