Stream Aligner: Efficient Sentence-Level Alignment via Distribution Induction

TL;DR

Stream Aligner introduces an efficient sentence-level alignment method that dynamically corrects LLM outputs, improving helpfulness, harmlessness, and math ability while reducing latency and reliance on larger models.

Contribution

It presents a novel alignment paradigm combining efficiency with enhanced performance through dynamic sentence correction using a small auxiliary model.

Findings

76.1% increase in helpfulness

36.0% increase in harmlessness

3.5% improvement in math ability

Abstract

The rapid advancement of large language models (LLMs) has led to significant improvements in their capabilities, but also to increased concerns about their alignment with human values and intentions. Current alignment strategies, including adaptive training and inference-time methods, have demonstrated potential in this area. However, these approaches still struggle to balance deployment complexity and capability across various tasks and difficulties. In this work, we introduce the Streaming Distribution Induce Aligner (Stream Aligner), a novel alignment paradigm that combines efficiency with enhanced performance in various tasks throughout the generation process. Stream Aligner achieves dynamic sentence-level correction by using a small model to learn the preferences of the suffix sentence, iteratively correcting the suffix sentence output by the upstream model, and then using the corrected sentence to replace the suffix sentence in subsequent generations. Compared to Aligner, our experiments demonstrate that Stream Aligner reduces reliance on the capabilities of additional models, enhances the reasoning abilities of LLMs, and decreases latency during user interaction. Specifically, Stream Aligner-2B model has achieved an improvement of 76.1% in helpfulness, 36.0% in harmlessness on the tested Llama2-70B-chat model, and Stream Aligner-8B has achieved an improvement of 3.5% on the math ability of the tested Llama3-70B-Instruct model.