Sonia Joseph
Scholar

Sonia Joseph

Google Scholar ID: UilqI20AAAAJ
Meta, McGill
multimodal modelsvideo modelsworld models
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
20
 
H-index
2
 
i10-index
0
 
Publications
10
 
Co-authors
0
 
Publications
11 items
What, Where, and How: Probing Spatiotemporal Representations in Video Foundation Models
2026
Cited
0
At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer Generalization
2026
Cited
0
Interpreting Physics in Video World Models
2026
Cited
0
Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learning
2026
Cited
0
Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval
2025
Cited
0
Into the Rabbit Hull: From Task-Relevant Concepts in DINO to Minkowski Geometry
2025
Cited
0
From Noise to Narrative: Tracing the Origins of Hallucinations in Transformers
2025
Cited
0
How Visual Representations Map to Language Feature Space in Multimodal LLMs
2025
Cited
0
Resume
Academic Achievements
  • Authored at least two articles related to artificial intelligence.
Research Experience
  • Published articles on the proper use of the logit lens; explored the topic of multimodal interpretability in 2024.
Background
  • Focused on AI research, particularly in the area of multimodal interpretability.