Paper ID: 2410.01103

Approximately Aligned Decoding

Daniel Melcer, Sujan Gonugondla, Pramuditha Perera, Haifeng Qian, Wen-Hao Chiang, Yanjun Wang, Nihal Jain, Pranav Garg, Xiaofei Ma, Anoop Deoras

It is common to reject undesired outputs of Large Language Models (LLMs); however, current methods to do so require an excessive amount of computation, or severely distort the distribution of outputs. We present a method to balance the distortion of the output distribution with computational efficiency, allowing for the generation of long sequences of text with difficult-to-satisfy constraints, with less amplification of low probability outputs compared to existing methods. We show through a series of experiments that the task-specific performance of our method is comparable to methods that do not distort the output distribution, while being much more computationally efficient.

Submitted: Oct 1, 2024