Dun Li Chan

                                                                                                                     
                                                                                                                     
`7MM"""Yb. `7MMF'   `7MF'`7MN.   `7MF'    `7MMF'      `7MMF'      .g8"""bgd `7MMF'  `7MMF'      db      `7MN.   `7MF'
  MM    `Yb. MM       M    MMN.    M        MM          MM      .dP'     `M   MM      MM       ;MM:       MMN.    M  
  MM     `Mb MM       M    M YMb   M        MM          MM      dM'       `   MM      MM      ,V^MM.      M YMb   M  
  MM      MM MM       M    M  `MN. M        MM          MM      MM            MMmmmmmmMM     ,M  `MM      M  `MN. M  
  MM     ,MP MM       M    M   `MM.M        MM      ,   MM      MM.           MM      MM     AbmmmqMA     M   `MM.M  
  MM    ,dP' YM.     ,M    M     YMM        MM     ,M   MM      `Mb.     ,'   MM      MM    A'     VML    M     YMM  
.JMMmmmdP'    `bmmmmd"'  .JML.    YM      .JMMmmmmMMM .JMML.      `"bmmmd'  .JMML.  .JMML..AMA.   .AMMA..JML.    YM

                                                    
                                                    
    .#*  :@*         *#    .@.         +#.   WW     
 M###@###@@###M   +@@@@@@@+*#.....      W:   WW     
 M*.:..WW..*.*#            M@@@@@W  .MMMMMM  WW     
 M*:@: +W :@::#    W@###@*.@.  WW    ****W#  WW     
 M* +M +W #* :#    W*   M*M@.  M*       .#*  WW     
 MW**+*MM*+**W#    WMWWW#W@#*  @:       MW   WW     
 MMWWWWWWWWWWM#    ++:::W:*+W :@       W#:.  WW     
 ..............    ++++++: :# +M      W@#+@* WW     
  :@########@*     ++++M@*  @:#*    .M#+# *@*WW     
  :#        M*        W#:   +##     WM.*#  : WW   +W
  :@MMMMMMMM@*    :++W@MM#* :@W     .  *#    WW   +M
  :#::::::::#*    *W++@*:: :##@*       *#    WW   WW
  :#********#*        @.  +@W W@+      *#    W#..:#+
  :@WWWWWWWW#*     :M## .#@+   +@M     *#    .#@@@M.
  .+        *.      **.  *.     :.     :+      ...

I work on the frontier of AI to uncover what lies within the latent space.

About Mebio Researchpapers Resumejuly 2026 GitHubcode LinkedInprofile Scholarcitations Contactemail

About Me

I'm currently building across AI safety evaluation in LLMs, mechanistic interpretability, and AI4Astro. My research interest lies in understanding how models work internally (layer-wise) and how this information can advance our progress in AI, while making sure they're safe to deploy in real-world settings.

Before that, I represented Malaysia at the International Olympiad on Astronomy and Astrophysics (IOAA) 2025 and the International Astronomy Olympiad (IAO) 2025; I also founded MYResearchGuide, a non-profit initiative to democratise science research opportunities for Malaysian youth.

Working Papers

Same Object, Different Telescope: A Layer-Wise Guide to Instrument and Physics Information. UniverseTBD with Kshitij Duraphe et al.

ongoing

Input Perturbations Propagate Differently: A Mechanistic Analysis of GPT-2 Robustness. with Emily Liu et al.

ongoing

Extended work of "Beyond Linearity in Attention Projections". with Marko Karbevski et al.

ongoing

Identification of Faint Near-Earth Asteroid Streak Candidates Using a Three-Channel Deep-Learning Pipeline. Drafting for journal submission; Dept. of Astronomy, Shanghai Jiao Tong University (SJTU) & University of Maryland with Yiyang Guo, Quanzhi Ye

ongoing

Feature-Based Comet Classification Pipeline for Zwicky Transient Facility (ZTF) Alert Image Streams. Dept. of Astronomy, University of Maryland with Quanzhi Ye et al.

ongoing

Publications (preprints/peer-reviewed papers)

Do All LLMs Know When They're Being Harmful? A Reproducibility Study of Latent-Space Safety Probes Across Model Families. EMNLP 9th BlackboxNLP Workshop Special Track: Reproducibility and Reliability in Interpretability Analyses (submitted) with Alizishaan Khatri

08/2026

Large-Scale Symbolic Music Genre Classification under Class Imbalance using Graph-Based MIDI Embeddings (part of GigaMIDI projects). Drafting for TISMIR with Keon Ju Maverick Lee, Philippe Pasquier et al., Metacreation Lab, Simon Fraser University (SFU)

07/2026

Anthropogenic Regional Adaptation in Multimodal Vision-Language Models. [arxiv] ACL ARR 2026 with Samuel Cahyawijaya, Peerat Limkonchotiwat et al., SEACrowd

04/2026

Can Linguistically Related Languages Guide LLM Translation in Low-Resource Settings? [official] [arxiv] EACL 2026 - LoResMT Workshop with Aishwarya Ramasethu, Niyathi Allu, Rohin Garg, Harshwardhan Fartale

03/2026

A Traditional Approach to Symbolic Piano Continuation. [arxiv] ISMIR 2025 - MIREX with Christian Zhou-Zheng, John Backsund, Alex Coventry, Avid Eslami, Jyotin Goel, Xingwen Han, Danysh Soomro, Galen Wei

09/2025