Ekdeep Singh Lubana
18 papers · 2021–2026 · 5 conferences · across top CS/AI conferences
Achievements
Jump to papers ↓+5 more ↓ Show less ↑
π Cross-Pollinator (4) π Interdisciplinary Bridge πΊοΈ Taxonomy Completionist (10) π§ Keyword Pioneer π Conference Polyglot (4)
π
Triple Crown
π€
Dynamic Duo
(12)
π
Century Club
(17)
β
The Questioner
(2)
β‘
Prolific Year
(7)
Conferences
ICLR (8)
ICML (5)
NIPS (3)
ACL (1)
NAACL (1)
Top co-authors
Research topics
Keywords
sparse autoencoder
(2)
neural network
(2)
representation learning
(2)
adversarial learning
(1)
preference alignment
(1)
preference optimization
(1)
direct preference optimization
(1)
loss landscape
(1)
matrix completion
(1)
concept learning
(1)
generative model
(1)
disentangled representation
(1)
learning dynamics
(1)
foundation model
(1)
hidden capability
(1)
training dynamics
(1)
mechanistic interpretability
(1)
neural network optimization
(1)
attention head
(1)
feature learning
(1)
Papers
From Isolation to Entanglement: When Do Interpretability Methods Identify and Disentangle Known Concepts?
ACL 2026
Swing-by Dynamics in Concept Learning and Compositional Generalization
ICLR 2025
ICLR: In-Context Learning of Representations
ICLR 2025
Competition Dynamics Shape Algorithmic Phases of In-Context Learning
ICLR 2025
A Percolation Model of Emergence: Analyzing Transformers Trained on a Formal Language
ICLR 2025
Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models
ICML 2025
Representation Shattering in Transformers: A Synthetic Study with Knowledge Editing
ICML 2025
Analyzing (In)Abilities of SAEs via Formal Languages
NAACL 2025
Compositional Capabilities of Autoregressive Transformers: A Study on Synthetic, Interpretable Tasks
ICML 2024
Abrupt Learning in Transformers: A Case Study on Matrix Completion
NIPS 2024
Emergence of Hidden Capabilities: Exploring Learning Dynamics in Concept Space
NIPS 2024
Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks
ICLR 2024
In-Context Learning Dynamics with Random Binary Sequences
ICLR 2024
Towards an Understanding of Stepwise Inference in Transformers: A Synthetic Graph Navigation Model
ICML 2024
What Makes and Breaks Safety Fine-tuning? A Mechanistic Study
NIPS 2024
Mechanistic Mode Connectivity
ICML 2023
What shapes the loss landscape of self supervised learning?
ICLR 2023
A Gradient Flow Framework For Analyzing Network Pruning
ICLR 2021