Cross-Tokenizer LLM Distillation through a Byte-Level Interface

TL;DR

This paper introduces Byte-Level Distillation (BLD), a simple method for cross-tokenizer knowledge transfer in language models that operates at the byte level, simplifying alignment and achieving competitive results.

Contribution

Proposes a novel byte-level interface for cross-tokenizer distillation, simplifying the process and outperforming complex heuristic methods on various benchmarks.

Findings

BLD performs competitively with sophisticated CTD methods.

Byte level serves as an effective common ground for knowledge transfer.

Consistent improvements across all tasks remain an open challenge.

Abstract

Cross-tokenizer distillation (CTD), the transfer of knowledge from a teacher to a student language model when the two use different tokenizers, remains a largely unsolved problem. Existing approaches rely on heuristic strategies to align mismatched vocabularies, introducing considerable complexity. In this paper, we propose a simple but effective baseline called Byte-Level Distillation (BLD) which enables CTD by operating at a common interface across tokenizers: the byte level. In more detail, we convert the teacher's output distribution to byte-level probabilities, attach a lightweight byte-level decoder head to the student, and distill through this shared byte-level interface. Despite its simplicity, BLD performs competitively with--and on several benchmarks surpasses--significantly more sophisticated CTD methods, across a range of distillation tasks with models from 1B to 8B parameters. Our results suggest that the byte level is a natural common ground for cross-tokenizer knowledge transfer, while also highlighting that consistent improvements across all tasks and benchmarks remain elusive, underscoring that CTD is still an open problem.