HotpotQA

Dataset Card for BEIR Benchmark hotpotqa is one of the datasets from the Question Answering task within BEIR, measuring Wikipedia article retrieval for a given multi-hop query. Dataset Summary BEIR is a heterogeneous benchmark built from 18 diverse datasets representing 9 information retrieval tasks. Fact-checking: FEV

Papers using HotpotQA (10)

StepOPSD: Step-Aware Online Preference Distillation for Agent Reinforcement Learning2026EMA Policy Gradient: Taming Reinforcement Learning for LLMs with EMA Anchor and Top-k KL2026Learning Query-Aware Budget-Tier Routing for Runtime Agent Memory2026A$^2$Search: Ambiguity-Aware Question Answering with Reinforcement Learning2025SAGE: Strategy-Adaptive Generation Engine for Query Rewriting2025Visual Agentic Reinforcement Fine-Tuning2025Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use2025AGILE: A Novel Reinforcement Learning Framework of LLM Agents2024 · 6 citesIM-RAG: Multi-Round Retrieval-Augmented Generation Through Learning Inner Monologues2024Enhancing Decision-Making for LLM Agents via Step-Level Q-Value Models2024