Predicting Rewards Alongside Tokens: Non-disruptive Parameter Insertion for Efficient Inference Intervention in Large Language Model Paper • 2408.10764 • Published Aug 20 • 8 • 2