Ubiquitous Knowledge Processing Lab

Project CARE

Code to reproduce the experimental results from the arXiv 2026 paper From Leaky Thoughts to Private Reasoning: Controlling What LRMs Say to Themselves

Murano: the interpretability research design framework.

Repository for the paper MASCS-Gen: Multi-Agent Synthetic Counseling Session Generation

A curated collection of papers, datasets, and resources on the topic of misleading chart understanding

FEGA, a pipeline for analyzing how sparse autoencoder features causally affect language-model behavior and representation geometry.

Plug-and-play implementation of Supervised Multi-Dimensional Scaling from the paper "Shape Happens: Automatic Feature Manifold Discovery in LLMs via Supervised Multi-Dimensional Scaling."