No One Fits All: From Fixed Prompting to Learned Routing in Multilingual LLMs

TL;DR

This paper evaluates various prompting strategies in multilingual large language models, revealing that no single approach is best for all languages and tasks, and proposes learned classifiers to optimize strategy selection.

Contribution

It introduces lightweight classifiers that predict optimal prompting strategies per instance, improving performance over fixed strategies and generalizing across unseen tasks.

Findings

Translation benefits low-resource languages despite imperfect quality.

No universal prompting strategy outperforms others across all languages.

Classifiers significantly improve strategy selection over fixed approaches.

Abstract

Translation-based prompting is widely used in multilingual LLMs, yet its effectiveness varies across languages and tasks. We evaluate prompting strategies across ten languages of different resource levels and four benchmarks. Our analysis shows that no single strategy is universally optimal. Translation strongly benefits low-resource languages even when translation quality is imperfect, high-resource languages gain little, and prompt-based self-routing underperforms explicit translation. Motivated by these findings, we formulate prompting strategy selection as a learned decision problem and introduce lightweight classifiers that predict whether native or translation-based prompting is optimal for each instance. The classifiers achieve statistically significant improvements over fixed strategies across four benchmarks and generalize to unseen task formats not observed during training. Further analysis reveals that language resource level, rather than translation quality alone, determines when translation is beneficial.