MSVAMP

The MSVAMP dataset/benchmark is used to evaluate the performance of models in cross-lingual reasoning tasks, particularly focusing on their effectiveness across both low-resource and high-resource languages.

Papers using MSVAMP (3)

TAPS: Task Aware Proposal Distributions for Speculative Sampling2026MERLIN: Multi-Stage Curriculum Alignment for Multilingual Encoder-LLM Integration in Cross-Lingual Reasoning2025Improving LLM Reasoning through Interpretable Role-Playing Steering2025