PPL
An Zhang
Papers (4)
Tiny Brains, Giant Impact: Uncovering the Keystone Neurons of LLM with Just a Few PromptsReasoning Can Be Restored by Correcting a Few Decision TokensExplaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional ExpansionSelf-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories