Wenya Wang

Multi-objective Large Language Model Alignment with Hierarchical Experts

Towards Generalizable Implicit In-Context Learning with Attention Routing

Adversarial Unlearning of Poisoned Features for Backdoor Defense in Federated Learning

CLUE: Conflict-guided Localization for LLM Unlearning Framework

Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective Mitigation

Knowledge Fusion of Large Language Models via Modular SkillPacks

Backdoor Attacks on Dense Retrieval via Public and Unintentional Triggers

Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates

Unveiling Language Skills under Circuits

On the Entropy of Language Models in Getting Semantic from Tokens