Wenqian Cui

PhD student at the Chinese University of Hong Kong. AI researcher in Music Generation and NLP.

[Shobhit Banga @shobhitbanga]

Every language in Monsoon V1, in one spin of the globe. 50 languages. 23 countries.

Spontaneous speech, not read prompts. If your ASR works in English and falls apart everywhere else, this is the data it was missing.

[Wenqian Cui @easy4u2remember]

@_WayWang @MeetResona Cool product! Feels like the realtime models aren’t just demos anymore!

[arXiv Sound @ArxivSound]

Wenqian Cui, Xiao-Hui Li, Daxin Tan, Qiyong Zheng, Irwin King, "Minimizing Modality Gap from the Input Side: Your Speech LLM Can Be a Prosody-Aware Text LLM," https://t.co/Dqc8v806gS

[Wenqian Cui @easy4u2remember]

@umpedronosapato Thanks man!

[Wenqian Cui @easy4u2remember]

Super excited to share that I recently had three papers accepted at AI conferences - two first-authored papers at ACL 2025 main conference and one at ICML 2025! 🎓🏆🚀 Immense gratitude to my brilliant co-authors who made this journey possible!

#NLP #ACL2025 #ICML2025

[Wenqian Cui @easy4u2remember]

I'm excited to continue pushing the boundaries of speech AI and would love to discuss research ideas in person! 🌟💡 If you're attending ACL 2025 in Vienna, let's meet up and explore potential collaborations. 🤝🌍

[Wenqian Cui @easy4u2remember]

3. "Improving Conversational Capabilities of Speech Language Models via Generative Dual-channel Spoken Dialogue Learning" (ICML 2025) 💬👥

[Wenqian Cui @easy4u2remember]

2. "VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Models" (ACL 2025) 🧪🎯  

ArXiv: https://t.co/wG3JgCCi8t
GitHub: https://t.co/hsYJzyo7On 📈🔬

[Wenqian Cui @easy4u2remember]

1. "Recent Advances in Speech Language Models: A Survey" (ACL 2025) 📚🔍  

ArXiv: https://t.co/XeZwxW2pdB
GitHub: https://t.co/W70T53yVMz 📊📝

[arXiv Sound @ArxivSound]

``VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Models,'' Wenqian Cui, Xiaoqi Jiao, Ziqiao Meng, Irwin King, https://t.co/yTkvaQFIKS

[Wenqian Cui @easy4u2remember]

9. ⚠ Current challenges in SpeechLM research include:
- Understanding component choices for better efficiency.
- Improving real-time generation capabilities.
- Addressing safety risks (toxicity & privacy).
There's much work ahead to enhance SpeechLMs! #ResearchChallenges #AI

[Wenqian Cui @easy4u2remember]

🚀 Really excited to share our new survey paper on Speech Language Models is now available! Dive deep into the world of speech-based foundation models and explore the future of human-computer interaction through voice.

Read it here: https://t.co/0owmqdvgVY
#SpeechLM #AI

[Wenqian Cui @easy4u2remember]

8. ✅ We categorize the evaluation of SpeechLMs into different aspects, including representation, linguistic, paralinguistic evaluations and more. We also emphasize the need for human evaluations of SpeechLMs.