
Bin Wang, Ph.D.
My career is driven by a dual passion: advancing technology for the benefit of humanity and the personal joy of discovery.
I am an AI researcher at MiroMind AI, an AGI-focused organization under Shanda Group, where I specialize in agentic model development. Before joining MiroMind, I conducted research at A*STAR and the National University of Singapore (NUS). I earned my PhD from the University of Southern California (USC).
I grew up in Jining, Shandong, and have since lived and worked in several cities across the globe, including Chengdu, Hong Kong, Ontario, and Los Angeles. I am currently based in Singapore, where I really enjoy staying.

Shot in 2019 Winter at Washington D.C.
Research
Starting from 2017, I entered the field of NLP and language and representation learning, which later became the core technique of AI after years. My early work focused on the intricacies of neural representations, a path that led to significant contributions in ACL, EMNLP, and NAACL.
As the field shifted toward massive pre-trained models, my research pivoted to address the challenges of Multimodal AI, specifically speech and audio signals, integrating linguistic depth with cross-domain data. Today, at MiroMind, I bridge the gap between "the lab" and "the product," applying my doctoral training and my years of research at A*STAR and NUS to solve the next generation of AI agentic challenges in the heart of Singapore's tech ecosystem.
- MiroThinker: Pushing the Performance Boundaries of Open-Source Research.. · Technical Report 2025
- MERaLiON-AudioLLM: Bridging Audio and Language with Large Language Models · ACL 2025
- AudioBench: A Universal Benchmark for Audio Large Language Models · NAACL 2025
- SeaEval for Multilingual Foundation Models: From Cross-Lingual.. · NAACL 2024
- Just Rank: Rethinking Evaluation with Word and Sentence Similarities · ACL 2022
- SBERT-WK: A Sentence Embedding Method by Dissecting BERT-based Word Models · TASLP 2020
- Evaluating Word Embedding Models: Methods and Experimental Results · ATSIP 2019
