About
Yiran Rex Ma
My name is MA Yiran (马义然), and I go by Yiran Rex Ma in papers. Call me Yiran or Rex per your liking :)
I’m with Peking University & ByteDance Digital Humanities Open Laboratory (PKUDH) and serving as a visiting senior algorithm engineer in Qwen Team, Alibaba Group, working on Multi-Modal Agents on cultural heritage.
Coming in September 2026, I’m starting my PhD in Theoretical and Applied Linguistics at Peking University, supervised by Prof. SU Qi. My concentration is on foundation language models, as well as computational linguistics, specifically on latent space of artificial neural networks.
Currently, I’m graduating from Beijing University of Posts and Telecommunications (BUPT), with a BA in English, Linguistics and Language Technologies and a minor in Data Science and Big Data Technologies. From September 2024 to August 2025, I was a research intern in AI for Humanities/Social Sciences subdivisions in TsinghuaNLP with Prof. LIU Zhiyuan. Before all that, I was primarily trained in the humanities with a discovered and ever since ongoing passion for human languages.
Current Scope
I’m striding towards being an LLM engineer, focusing on pretraining and specifically, extra-long sequence training.
Delicate, hand-crafted context/harness engineering is, to me, only a temporary workaround, giving everyone a false sense of “control” while tiring us out.
What I’m truly after is to build native capabilities into the model to handle long contexts (e.g., responsible and flexible attention) with minimal performance degradation, to reproduce the scaling law in contexts, and to implement all of this in various configurations and real-world scenarios.
Also, SSMs and Diffusion LLMs seem cool. Should we blend them in?
Last but not least
I’m something of a language aficionado, ever since I was made to realize this when my education finally took English seriously at the age of 13. Unlike many of my peers in this context, my family is nothing but humble small town dwellers from the rurals, with nothing fancy but unspoken, Asian-style support.
Therefore, American English and Simplified Chinese are my primary working languages, while I was born and raised with Sichuan Dialect as my mother tone and Mandarin Chinese my first language. Besides, I’m currently on track for Spanish, French, and Traditional Chinese (and more…).
And if we think of language (both natural and programming) as a human-driven probability system, I also hope to establish proficiency bestriding among major languages, to become a (self-proclaimed) full-stack polyglot :)
Oh, almost forgot, I also produce my own music under the name RexEra, experimenting with all sorts of elements while documenting my life.
This is an academic/personal profile/blog, where I update news and whimsical thoughts. Enjoy!