Search papers, labs, and topics across Lattice.
This paper introduces UL-XCoT, a framework for efficient cross-lingual chain-of-thought reasoning that reduces computational cost by selecting a small set of relevant languages in a unified logic space and pruning low-quality reasoning paths during decoding. UL-XCoT minimizes token usage and latency, achieving greater efficiency under limited sampling budgets. Experiments on PolyMath and MMLU-ProX-Lite demonstrate that UL-XCoT achieves competitive accuracy while cutting over 50% of decoding token cost compared to existing sampling baselines, especially for low-resource languages.
Reasoning across languages doesn't have to break the bank: a new framework slashes token costs by over 50% while maintaining accuracy, especially boosting performance in low-resource languages.
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.
Tsinghua AI