Less Languages, Less Tokens: An Efficient Unified Logic Cross-lingual Chain-of-Thought Reasoning Framework

TL;DR

UL-XCoT is a novel framework that improves multilingual reasoning efficiency by reducing language and token redundancy, leading to faster inference and better robustness across diverse languages.

Contribution

It introduces a unified logic space and dynamic pruning techniques to significantly cut decoding costs while maintaining competitive accuracy in multilingual reasoning.

Findings

Achieves over 50% reduction in decoding token cost.

Demonstrates stable improvements on low-resource languages.

Maintains competitive accuracy across multiple multilingual benchmarks.

Abstract

Cross-lingual chain-of-thought (XCoT) with self-consistency markedly enhances multilingual reasoning, yet existing methods remain costly due to extensive sampling of full trajectories across languages. Moreover, multilingual LLM representations vary strongly by language, hindering direct feature comparisons and effective pruning. Motivated by this, we introduce UL-XCoT, the first efficient unified logic cross-lingual reasoning framework that minimizes redundancy in token usage and latency, yielding the greatest efficiency under limited sampling budgets during inference. Specifically, UL-XCoT (1) achieves less languages by selecting, per query, a small candidate language set in a language-invariant unified logic space, (2) enables less tokens by monitoring logic-space trajectory dynamics during decoding to prune low-quality reasoning paths, and (3) aggregates the remaining high-quality trajectories via voting. Experiments on PolyMath across 18 languages and MMLU-ProX-Lite across 29 languages with DeepSeek-R1-DistillQwen-7B demonstrate that UL-XCoT achieves competitive accuracy while sharply cutting over 50% decoding token cost versus prior sampling baselines. UL-XCoT also delivers more stable gains on low-resource languages, underscoring consistently superior robustness where standard XCoT self-consistency method fails.