Tutorials
ReasoningBank: Enabling agents to learn from experience
ReasoningBank is a memory framework that distills generalizable reasoning strategies from an agent's successful and failed experiences, which the agent retrieves at test time to inform interactions wh
ReasoningBank is a memory framework that distills generalizable reasoning strategies from an agent's successful and failed experiences, which the agent retrieves at test time to inform interactions while integrating new learnings back into memory . Combined with memory-aware test-time scaling (MaTTS), the framework accelerates learning by scaling up the agent's interaction experience . This approach delivers up to 34.2% relative effectiveness gains and 16% fewer interaction steps compared to prior memory designs across web and software-engineering benchmarks .
Related
- Coding agents learn from experience, but that knowledge stays locked in silos. Solve a thousand SWE tasks, and none of that wisdom helps wit…
- From Procedural Skills to Strategy Genes: Towards Experience-Driven Test-Time Evolution
- Learning to Forget -- Hierarchical Episodic Memory for Lifelong Robot Deployment
- TRACE: An Experiential Framework for Coherent Multi-hop Knowledge Graph Question Answering
Source: Google Research | 2026-04-21