Jiachen Zhu
Scholar

Jiachen Zhu

Google Scholar ID: sao1OhsAAAAJ
PhD Student, New York University
Self-Supervised LearningComputer Vision
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
343
 
H-index
7
 
i10-index
7
 
Publications
7
 
Co-authors
0
 
Publications
2 items
Small LLMs: Pruning vs. Training from Scratch
2026
Cited
0
The Spike, the Sparse and the Sink: Anatomy of Massive Activations and Attention Sinks
2026
Cited
0
Resume
Academic Achievements
  • Papers:
  • - Scaling Language-Free Visual Representation Learning
  • - Transformers without Normalization
  • - MetaMorph: Multimodal Understanding and Generation via Instruction Tuning
  • - Variance-Covariance Regularization Improves Representation Learning
  • - VoLTA: Vision-Language Transformer with Weakly-Supervised Local-Feature Alignment
  • - Masked Siamese ConvNets
  • - TiCo: Transformation Invariance and Covariance Contrast for Self-Supervised Visual Representation Learning
Research Experience
  • Currently a fifth-year PhD candidate in Computer Science at NYU Courant. Visiting Researcher at FAIR, Meta, hosted by Zhuang Liu and Koustuv Sinha.
Education
  • PhD, Computer Science, New York University, 2020 - Now; Advisor: Yann LeCun
  • MSc, Computer Science, New York University, 2018 - 2020
  • BSc, Computer Science, The Hong Kong Polytechnic University, 2010 - 2015
Background
  • Research interests: self-supervised learning for images and videos, as well as pretraining vision encoders for vision-language models (VLMs). Also interested in understanding the design of all kinds of neural network architectures.
Miscellany
  • Contact: jiachen DOT zhu AT nyu DOT edu
  • My Favourite Illusion: 1, 2