This paper investigates the effectiveness of in-context learning (ICL) and cross-lingual in-context learning (X-ICL) for low-resource languages. The study evaluates ICL and X-ICL on 25 low-resource and 7 higher-resource languages, focusing on the impact of in-context alignment strategies. The results show that in-context label alignment is not effective for most languages, while in-context query alignment significantly improves performance. The study also highlights the importance of semantic similarity and cross-lingual retrieval in enhancing the performance of LLMs on low-resource languages. The findings suggest that X-ICL can be a viable alternative to traditional methods when there is no available corpus for the specific task. The paper concludes that in-context query alignment is more effective than in-context label alignment for low-resource languages and that cross-lingual semantic similarity is crucial for improving the performance of LLMs on low-resource languages. The study also emphasizes the importance of format consistency in improving the performance of LLMs on low-resource languages. The results indicate that X-ICL is less effective than translate-test ICL for low-resource languages, but it remains a valuable approach when no machine translation model is available. The paper provides insights into the challenges of applying LLMs to low-resource languages and suggests that future research should focus on improving the generalization capabilities of LLMs for these languages. The study also highlights the importance of using parallel data and cross-lingual semantic similarity in improving the performance of LLMs on low-resource languages. The paper concludes that in-context query alignment is a more effective approach than in-context label alignment for low-resource languages and that cross-lingual semantic similarity is crucial for improving the performance of LLMs on low-resource languages. The study also emphasizes the importance of format consistency in improving the performance of LLMs on low-resource languages. The results indicate that X-ICL is less effective than translate-test ICL for low-resource languages, but it remains a valuable approach when no machine translation model is available. The paper provides insights into the challenges of applying LLMs to low-resource languages and suggests that future research should focus on improving the generalization capabilities of LLMs for these languages.This paper investigates the effectiveness of in-context learning (ICL) and cross-lingual in-context learning (X-ICL) for low-resource languages. The study evaluates ICL and X-ICL on 25 low-resource and 7 higher-resource languages, focusing on the impact of in-context alignment strategies. The results show that in-context label alignment is not effective for most languages, while in-context query alignment significantly improves performance. The study also highlights the importance of semantic similarity and cross-lingual retrieval in enhancing the performance of LLMs on low-resource languages. The findings suggest that X-ICL can be a viable alternative to traditional methods when there is no available corpus for the specific task. The paper concludes that in-context query alignment is more effective than in-context label alignment for low-resource languages and that cross-lingual semantic similarity is crucial for improving the performance of LLMs on low-resource languages. The study also emphasizes the importance of format consistency in improving the performance of LLMs on low-resource languages. The results indicate that X-ICL is less effective than translate-test ICL for low-resource languages, but it remains a valuable approach when no machine translation model is available. The paper provides insights into the challenges of applying LLMs to low-resource languages and suggests that future research should focus on improving the generalization capabilities of LLMs for these languages. The study also highlights the importance of using parallel data and cross-lingual semantic similarity in improving the performance of LLMs on low-resource languages. The paper concludes that in-context query alignment is a more effective approach than in-context label alignment for low-resource languages and that cross-lingual semantic similarity is crucial for improving the performance of LLMs on low-resource languages. The study also emphasizes the importance of format consistency in improving the performance of LLMs on low-resource languages. The results indicate that X-ICL is less effective than translate-test ICL for low-resource languages, but it remains a valuable approach when no machine translation model is available. The paper provides insights into the challenges of applying LLMs to low-resource languages and suggests that future research should focus on improving the generalization capabilities of LLMs for these languages.