Qiaolin Wang

Building Audio General Intelligence
Researcher at Bland | qw2443@columbia.edu

I am Qiaolin Wang, a researcher at Bland, where I work on expressive ASR, TTS, post-training, and speech-to-speech systems. I received my M.S. in Electrical Engineering from Columbia University, advised by Professor Nima Mesgarani.

I am passionate about building models that can Perceive, Reason, and Speak as naturally as humans do. My research has examined the fundamental capabilities of Large Audio Language Models (LALMs): their internal representations of syntax and context (EMNLP 2025, SAC Highlight), their capacity for cross-modal reasoning (ICASSP 2026), and their contextual and cultural understanding of audio-visual media (COLM 2026).

Before Columbia, I earned my B.Eng. in Computer Science from Wuhan University and was a research intern at Wiz.AI, where I built a state-of-the-art speech emotion recognition LLM.

News

Sep 28, 2026 📄 NVAlign is now on arXiv! Direct-gradient post-training that improves non-verbal control in continuous autoregressive flow-matching TTS. Audio samples here.
Jul 08, 2026 🎉 AVMeme Exam was accepted to COLM 2026! Explore the benchmark.
Apr 06, 2026 🚀 I joined Bland as a researcher, working on expressive ASR, TTS, post-training, and speech-to-speech systems.
Jan 25, 2026 📄 AVMeme Exam is now on arXiv! A multimodal, multilingual, multicultural benchmark for LLMs’ contextual and cultural knowledge. Project page here.
Jan 17, 2026 🎉 SightSound-R1 was accepted to ICASSP 2026! Read it on IEEE Xplore or arXiv.
Nov 10, 2025 🏆 Our paper was awarded EMNLP SAC Highlight!
Nov 07, 2025 🎤 I will be presenting our work at the SANE 2025 workshop at Google NYC!
Sep 19, 2025 📄 SightSound-R1 is now available on arXiv! In this work, we propose a framework to transfer reasoning from VLMs to LALMs.
Aug 07, 2025 🎉 Thrilled that our paper was accepted for an oral presentation at EMNLP 2025!

Selected Publications

  1. arXiv
    nvalign.png
    NVAlign: Direct-Gradient Optimization for Non-Verbal Control in Continuous Autoregressive Flow Matching Text-to-Speech
    Qiaolin Wang, Pedro Sandoval-Segura, Anunaya Joshi, and 2 more authors
    arXiv preprint arXiv:2609.31892, 2026
    Submitted to ICASSP 2027
    arXiv PDF Website
  2. COLM
    avmeme_exam_stacked.png
    AVMeme Exam: A Multimodal Multilingual Multicultural Benchmark for LLMs’ Contextual and Cultural Knowledge and Thinking
    Xilin Jiang, Qiaolin Wang, Junkai Wu, and 30 more authors
    In Conference on Language Modeling (COLM), 2026
    arXiv PDF Website
  3. EMNLP
    layerwise_probing.jpg
    Layer-wise Minimal Pair Probing Reveals Contextual Grammatical-Conceptual Hierarchy in Speech Representations
    Linyang He*, Qiaolin Wang*, Xilin Jiang, and 1 more author
    In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025
    Oral presentation · SAC Highlight
    PDF
  4. ICASSP
    sightsound_r1.png
    SightSound-R1: Cross-Modal Reasoning Distillation from Vision to Audio Language Models
    Qiaolin Wang, Xilin Jiang, Linyang He, and 2 more authors
    In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026
    DOI arXiv HTML PDF

⭐ GitHub Stars in 3D

A rotating 3D galaxy with one particle for every GitHub star my repositories have earned A 3D city of GitHub stars per quarter and repository, with the running total on the back wall

GitHub stars GitHub followers

🎬 Where it started

My speech AI journey began as a side quest in 2022: teaching a model to speak as Paimon from Genshin Impact.

  • Built and annotated a multi-speaker dataset of ≈48,000 clips (15 h) from 50 Genshin Impact characters, using ECAPA-TDNN for speaker classification and Whisper for transcription
  • Fine-tuned a VITS speech synthesis model on a curated set of Paimon clips
  • Released it as a technical demo on Bilibili (600K+ views) with a public Colab for anyone to try