Raya Horesh

MEAL: A Multi-dimensional Evaluation of Alignment Techniques for LLMs

An Efficient and Modular Framework for Targeted Harm Mitigation in LLMs

Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding

Out-of-Distribution Detection using Synthetic Data Generation

Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding

Out-of-Distribution Detection using Synthetic Data Generation

SPRI: Aligning Large Language Models with Context-Situated Principles