Bingzhe Wu

SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs via Stable Safety-critical Expert Identification

NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Prediction

Measuring Diversity in Synthetic Datasets

IBCircuit: Towards Holistic Circuit Discovery with Information Bottleneck

Evaluating Diversity of LLM-generated Datasets: A Classification Perspective

IBCircuit: Towards Holistic Circuit Discovery with Information Bottleneck

Parrot: Seamless Spoken Dialogue Interaction with Double-Channel Large Language Models

VDC: Versatile Data Cleanser based on Visual-Linguistic Inconsistency by Multimodal Large Language Models

Language Agents for Detecting Implicit Stereotypes in Text-to-image Models at Scale

Adapting Large Language Models for Content Moderation: Pitfalls in Data Engineering and Supervised Fine-tuning