Dhanya Sridhar

Assistant Professor@Université de Montréal and Mila-Quebec AI Institute·加拿大·OpenReview

研究方向

probabilistic models · variational inference · bayesian models · text analysis · topic models · algorithmic fairness · deep generative models · causality · causal inference · probabilistic graphical models

5.0

On the Identifiability of Concepts from Large Language Model Activations

ICLR 2026Rejected

通讯

3.5

From Isolation to Entanglement: When Do Interpretability Methods Identify and Disentangle Known Concepts?

ICLR 2026Withdrawn

7.3

Causal Differentiating Concepts: Interpreting LM Behavior via Causal Representation Learning

NeurIPS 2025Spotlight

通讯

6.0

Does learning the right latent variables necessarily improve in-context learning?

ICML 2025Poster

通讯

5.6

Does learning the right latent variables necessarily improve in-context learning?

合作者 (20)

Dhanya Sridhar

On the Identifiability of Concepts from Large Language Model Activations

From Isolation to Entanglement: When Do Interpretability Methods Identify and Disentangle Known Concepts?

Causal Differentiating Concepts: Interpreting LM Behavior via Causal Representation Learning

Does learning the right latent variables necessarily improve in-context learning?

Does learning the right latent variables necessarily improve in-context learning?

In-context learning and Occam's razor

Demystifying amortized causal discovery with transformers

In-Context Learning and Occam's Razor