ScienceWorld

ScienceWorld is a benchmark dataset used to evaluate the performance and generalization of large language model agents in complex interactive decision-making tasks.

Papers using ScienceWorld (13)

RLVMR: Reinforcement Learning with Verifiable Meta-Reasoning Rewards for Robust Long-Horizon Agents2025 · 35 citesRetrospex: Language Agent Meets Offline Reinforcement Learning Critic2025 · 7 citesBeyond Policy Optimization: A Data Curation Flywheel for Sparse-Reward Long-Horizon Planning2025 · 3 citesPolicy-Conditioned Counterfactual Credit for Verifiable Reinforcement Learning of Long-Horizon Language Agents2026Milestone-Guided Policy Learning for Long-Horizon Language Agents2026Hierarchical Reinforcement Learning with Augmented Step-Level Transitions for LLM Agents2026Think Fast and Slow: Step-Level Cognitive Depth Adaptation for LLM Agents2026Exploratory Memory-Augmented LLM Agent via Hybrid On- and Off-Policy Optimization2026Differentiable Evolutionary Reinforcement Learning2025EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learning2025Reward Is Enough: LLMs Are In-Context Reinforcement Learners2025Unleashing Embodied Task Planning Ability in LLMs via Reinforcement Learning2025Divide and Conquer: Grounding LLMs as Efficient Decision-Making Agents via Offline Hierarchical Reinforcement Learning2025