SPADE
Self-play RL framework where LLMs design environments and solve them for continuous improvement
About
Self-Play in Adaptive Synthetic Executable Environments (SPADE) is a research framework that enables LLMs to self-improve through dual-role playing. A single model acts as both Environment Designer (creating increasingly complex tasks) and Agent (solving them), forming a self-play loop for continuous learning without external supervision. Published by Bo Liu et al., it demonstrates how agents can bootstrap their capabilities through adaptive curriculum generation, similar to how AlphaGo improved through self-play but applied to general task completion.
Details
| Language | |
| Patterns |
Tags
Quick Info
- Organization
- Research (Bo Liu et al.)
- Pricing
- open-source
- Free Tier
- Yes
- Updated
- Aug 20, 2026
Also in Frameworks
LangChain
Build context-aware reasoning applications with LLMs
AutoGen
Microsoft's framework for building multi-agent AI systems
CrewAI
Multi-agent orchestration framework for collaborative AI workflows