Researchers release paper on Steward, a multi-agent LLM autonomous investment system
Get up to speed
Researchers including Yebo Feng have released a research paper on Steward, a multi-agent LLM autonomous investment system that defines self-evolution boundaries through a type system, operating around an immutable core that cannot be self-edited. The system's core includes nine role-specialized agents, seven layers of shared memory, and a verifiable quarterly reflective self-evolution mechanism. The paper states that during a five-month live deployment (102 trading days, 1632 LLM calls), the system controlled a 24.6% sector drawdown to a 0.3% loss. Backtests from 2015 to 2026 show that the core alone reduces maximum drawdown across all 50 random entry strategies.
AI-generated from reports · 1 hr agoUpdated
Coverage timeline
Follow the coverage to understand different sides of the event.
- arXiv q-finarXiv paper proposes Steward, a multi-agent LLM autonomous investment system with an invariant core
This paper proposes Steward, a multi-agent LLM autonomous investment system that defines self-evolution boundaries through a type system. Its core includes nine types of role-specialized agents, seven layers of shared memory, and a verifiable quarterly reflective self-evolution mechanism, all operating around an immutable core that cannot be self-edited.
Trend of this event
Not enough consecutive observations yet; trend is not drawn.