MuSiQue

The 'MuSiQue' dataset is a benchmark used to evaluate open-domain question answering systems, specifically focusing on questions that can have multiple valid answers.

Papers using MuSiQue (4)

SEM: Reinforcement Learning for Search-Efficient Large Language Models2025 · 8 citesA$^2$Search: Ambiguity-Aware Question Answering with Reinforcement Learning2025Fathom-DeepResearch: Unlocking Long Horizon Information Retrieval and Synthesis for SLMs2025Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use2025