Hi! I'm Caden. I'm currently thinking about interfaces for interpretability and (occasionally) chasing some research threads at a nonprofit AI lab.

Email: kh4dien [at] gmail [dot] com

Find me on Github | Twitter

Current

• Bau Lab
• NNsight

Work

• Steering Fine Tuning with Targeted Concept Ablation
  Helena Casademunt*, Caden Juang*, Samuel Marks, Senthooran Rajamanoharan, Neel Nanda
  ICLR SLLM, ICLR BuildingTrust
  paper | code
• Sparsifying a Model's Computations: Preliminary Findings
  Jacob Drori, Jannik Brinkmann, Logan Riggs, Caden Juang
  LessWrong
  post
• Robustly identifying concepts introduced during chat fine-tuning using crosscoders
  Julian Minder*, Clement Dumas*, Caden Juang, Bilal Chughtai, Neel Nanda
  Arxiv
  paper | code
• Automatically Interpreting Millions of Features in Large Language Models
  Gonçalo Paulo, Alex Mallen, Caden Juang, Nora Belrose
  ICML 2025
  paper | code
• Open Source Automated Interpretability for Sparse Autoencoder Features
  Caden Juang*, Gonçalo Paulo*, Jacob Drori, Nora Belrose
  Eleuther
  post | code | demo
• NNsight and NDIF: Democratizing Access to Foundation Model Internals
  Jaden Fiotto-Kaufman*, Alexander R Loftus*, Eric Todd, Jannik Brinkmann, Caden Juang, Koyena Pal, Can Rager, Aaron Mueller, Samuel Marks, Arnab Sen Sharma, Francesca Lucchetti, Michael Ripa, Adam Belfki, Nikhil Prakash, Sumeet Multani, Carla Brodley, Arjun Guha, Jonathan Bell, Byron Wallace, David Bau
  ICLR 2025
  paper | code
• Poser: Unmasking Alignment Faking LLMs by Manipulating Their Internals
  Joshua Clymer*, Caden Juang*, Severin Field*
  Arxiv
  paper

Misc

• MATS w Neel Nanda
• Autointerp
• NNterface
• MAIA - student group
• GammaTau - PL and AI group
• ARENA 3.0
• Beaverworks
• Spotify
• M3

---

Last updated: 6/2025 | source