GPQA

Dataset Card for GPQA GPQA is a multiple-choice, Q&A dataset of very hard questions written and validated by experts in biology, physics, and chemistry. When attempting questions out of their own domain (e.g., a physicist answers a chemistry question), these experts get only 34% accuracy, despite spending >30m with ful

Papers using GPQA (20)

Stable Reinforcement Learning for Efficient Reasoning2025 · 25 citesRUMAD: Reinforcement-Unifying Multi-Agent Debate2026 · 2 citesRL with Learnable Textual Feedback: A Bilevel Approach2026Learning to Orchestrate Agents in Natural Language with the Conductor2025Adaptive Graph of Thoughts: Test-Time Adaptive Reasoning Unifying Chain, Tree, and Graph Structures2025 · 1 citesApriel-1.5-OpenReasoner: RL Post-Training for General-Purpose and Efficient Reasoning2026Off-Policy Value-Based Reinforcement Learning for Large Language Models2026CLEANER: Self-Purified Trajectories Boost Agentic Reinforcement Learning2026Can LLMs Guide Their Own Exploration? Gradient-Guided Reinforcement Learning for LLM Reasoning2025Online Rubrics Elicitation from Pairwise Comparisons2025Reasoning with Sampling: Your Base Model is Smarter Than You Think2025Train for Truth, Keep the Skills: Binary Retrieval-Augmented Reward Mitigates Hallucinations2025Sample More to Think Less: Group Filtered Policy Optimization for Concise Reasoning2025RL of Thoughts: Navigating LLM Reasoning with Inference-time Reinforcement Learning2025SHARP: Synthesizing High-quality Aligned Reasoning Problems for Large Reasoning Models Reinforcement Learning2025Enigmata: Scaling Logical Reasoning in Large Language Models with Synthetic Verifiable Puzzles2025Reinforcing General Reasoning without Verifiers2025Maximizing Confidence Alone Improves Reasoning2025Synthetic Data RL: Task Definition Is All You Need2025CPL: Critical Plan Step Learning Boosts LLM Generalization in Reasoning Tasks2024