Multi-Granularity Semantic Revision for Large Language Model Distillation

TL;DR

This paper introduces a multi-granularity semantic revision approach for large language model distillation, addressing errors and misalignments in output by correcting sequences, adaptively focusing on semantic-rich outputs, and enforcing span-level probability consistency.

Contribution

It proposes a novel multi-level semantic revision framework incorporating sequence correction, adaptive loss, and span correlation constraints for improved LLM distillation.

Findings

Outperforms existing distillation methods across various model sizes.

Reduces generation errors and enhances semantic transfer.

Demonstrates effectiveness on models from 0.1B to 13B parameters.

Abstract

Knowledge distillation plays a key role in compressing the Large Language Models (LLMs), which boosts a small-size student model under large teacher models' guidance. However, existing LLM distillation methods overly rely on student-generated outputs, which may introduce generation errors and misguide the distillation process. Moreover, the distillation loss functions introduced in previous art struggle to align the most informative part due to the complex distribution of LLMs' outputs. To address these problems, we propose a multi-granularity semantic revision method for LLM distillation. At the sequence level, we propose a sequence correction and re-generation (SCRG) strategy. SCRG first calculates the semantic cognitive difference between the teacher and student to detect the error token, then corrects it with the teacher-generated one, and re-generates the sequence to reduce generation errors and enhance generation diversity. At the token level, we design a distribution adaptive clipping Kullback-Leibler (DAC-KL) loss as the distillation objective function. DAC-KL loss exploits a learnable sub-network to adaptively extract semantically dense areas from the teacher's output, avoiding the interference of redundant information in the distillation process. Finally, at the span level, we leverage the span priors of a sequence to compute the probability correlations within spans, and constrain the teacher and student's probability correlations to be consistent, further enhancing the transfer of semantic information. Extensive experiments across different model families with parameters ranging from 0.1B to 13B demonstrate the superiority of our method compared to existing methods.