AgentGym
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| One-step next-observation prediction | AgentGym Unweighted Average (test) | Token F185 | 16 | |
| Agentic Reasoning and Interaction | AgentGym | ALF Success Rate88 | 11 | |
| Interactive decision-making | AgentGym WebShop, ALFWorld, BabyAI (test) | WebShop Success Rate78.8 | 8 | |
| Language Agent Task | AgentGym All (test) | Success Rate69.3 | 5 | |
| Language Agent Task | AgentGym ALFWorld (test) | Success Rate36 | 5 | |
| Language Agent Task | AgentGym SciWorld (test) | Success Rate91.2 | 5 |
Showing 6 of 6 rows