SPADE Paper: Single Model Self-Implements Environment Design and Agent Self-Play

_AndrewZhao · x · 2026-08-21

Natasha Jaques et al. release the SPADE paper, proposing recursive self-improvement via multi-agent RL and Unsupervised Environment Design (UED). The framework uses a single LLM to act as both an "Environment Designer" (building multi-turn RL training environments using the Gym step()/reset() API) and a "Reasoning Agent" that learns to solve them. The Designer optimizes task difficulty by maximizing the Agent's regret, achieving automatic scaling of training environments.

Related event: SPADE: Single-Model Self-Play Generates Ever-Harder RL Environments for Continued Self-Improvement(7 posts)→

Original post →

More from Research

• MIT uses ML to screen catalysts for greener ammonia production — nordicinst · 2026-08-21
• Understanding models' reasons: a research agenda on AI behavior — brwilder · 2026-08-21
• First Computer-Use Dataset for Design Open Sourced: 3400+ Real Figma Trajectories — CShorten30 · 2026-08-21
• Monroe: MFM for In-Context Probabilistic Inference — chaumian · 2026-08-21
• Geoffrey Irving: Verified Lean kernel enables wild, sketchy optimizations for speed — geoffreyirving · 2026-08-21
• Scaling expert supervision is the bottleneck in frontier data, says SnorkelAI expert — ShayneRedford · 2026-08-21

Research channel →