Chandan Singh

Eliciting and evaluating generalizable explanations from large reasoning models

Mixture of Inputs: Text Generation Beyond Discrete Token Sampling

Interpretable Next-token Prediction via the Generalized Induction Head

Bayesian Concept Bottleneck Models with LLM Priors

Interpretable Language Modeling via Induction-head Ngram Models

Simplifying DINO via Coding Rate Regularization

Vector-ICL: In-context Learning with Continuous Vector Representations

Model Tells Itself Where to Attend: Steerable Prompting for Reliable Reading Comprehension of LLM

Evaluating Graphical Perception of Large Multimodal Models