Sokoban
Emerging7papers using it
2019first seen
Sokoban is a benchmark dataset used to evaluate the performance of agents in solving spatial reasoning and planning tasks involving pushing boxes to designated locations within a grid-based environment.
Papers using Sokoban (5)
- Retrospective Progress-Aware Self-Refinement for LLM Agent TrainingTAPE: Tool-guided Adaptive Planning And Constrained Execution In Language Model AgentsHiMAC: Hierarchical Macro-Micro Learning for Long-Horizon LLM AgentsTSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM AgentsMeta-RL Induces Exploration in Language Agents