Hila Chefer
Scholar

Hila Chefer

Google Scholar ID: B8sA9JoAAAAJ
PhD student at Tel Aviv University, Meta
Deep LearningComputer VisionNLPExplainable AI
Citations & Impact
All-time
Citations
2,800
 
H-index
11
 
i10-index
11
 
Publications
14
 
Co-authors
16
list available
Resume (English only)
Academic Achievements
  • Work covered by The Verge, ZDNET, Two Minute Papers, Analytics India Magazine, etc.
  • Pioneered Transformer explainability methods for single and multi-modality
  • Developed Attend-and-Excite: inference-time semantic guidance for diffusion models
  • Co-developed Lumiere: a foundation model for efficient text-to-video generation (with Google Research)
  • Led VideoJAM: a novel training framework achieving SOTA motion and physics understanding in video generation (with Meta AI)
  • Two papers accepted to NeurIPS 2025 (FlowMo and Revisiting LRP)
  • VideoJAM accepted as oral presentation at ICML'25 (top ~1%)
  • Delivering two keynotes and co-organizing two workshops at ICCV'25