CharED: Character-wise Ensemble Decoding for Large Language Models

Kevin Gu, Eva Tuecke, Dmitriy A Katz, Raya Horesh, David Alvarez-Melis, Mikhail Yurochkin

Keywords: large language models, ensembling

TL;DR: We introduce CharED, a method for “averaging” outputs from multiple models even for LLMs with different vocabularies and tokenizers, by converting subword-level LLMs into character-level ones at the decoding step.

Abstract: Large language models (LLMs) have shown remarkable potential for problem solving, with open source models achieving increasingly impressive performance on benchmarks measuring areas from logical reasoning to mathematical ability. Ensembling models can further improve capabilities across a variety of domains. However, conventional methods of combining models at inference time such as shallow fusion necessitate a shared vocabulary and tokenization, and alternatives like fine-tuning for domain-specific performance are both time consuming and computationally expensive. We therefore present an inference-time ensembling algorithm aimed at ``averaging'' outputs from multiple LLMs and illustrate its improved performance across multiple domains compared to its constituent models alone. Character-wise ensemble decoding (CharED) finds the marginal distribution of each character for an individual model and performs a weighted average to generate an output, character by character. In coding, math, and toxicity benchmarks, we find our proposed model able to combine complementary strengths of multiple LLMs, regardless of vocabulary, tokenization, or model size.

Submission Number: 67