Portrait of HongYu Liu

Speech Language Model · SLM Trustworthiness · Agentic RL

HongYu Liu

I build trustworthy speech and language systems, with research spanning proactive interaction, privacy evaluation, and end-to-end spoken dialogue intelligence.

Current Focus:

Education

Bachelor's Degree

South China Normal University

B.Eng. in Software Engineering

Overall GPA: 4.06.

Matriculation: Sep 2026

The Chinese University of Hong Kong, Shenzhen

Master of Science in Data Science

Enrollment scheduled for September 2026.

Internship

Feb 2026 - Present

Insta360 · Speech Algorithm Intern

Shenzhen, China

Developing production-grade speech algorithms and optimizing low-latency model inference for voice applications.

Jun 2025 - Sep 2025

Amphion Technology · R&D Intern

Shenzhen, China

Developed core algorithms for a voice-cloning application and supported backend model integration for video translation.

Research

ICLR 2026 · Poster

VoxPrivacy: A Benchmark for Evaluating Interactional Privacy of Speech Language Models

Introduced a multi-user benchmark for measuring interactional privacy risks in speech-language models; second author.

ECAI 2025 · Oral

DialogGraph-LLM: Graph-Informed LLMs for End-to-End Audio Dialogue Intent Recognition

Proposed a graph-informed framework for end-to-end intent recognition in spoken dialogue; first author.

ADMA 2025 · Poster

Multi-segment Multitask Fusion Network for Marketing Audio Classification

Proposed MSMT-FN for marketing-call attitude classification, achieving stronger results than prior baselines; first author.