Trade in Minutes! Rationality-Driven Agentic System for Quantitative Financial Trading
Researchers present TiMi (Trade in Minutes), a multi-agent financial trading system powered by large language models. TiMi separates strategy design from rapid trading, using macro market analysis and micro pair-specific customization. In live tests on over 200 stock and crypto pairs, TiMi beat traditional and machine learning methods with higher returns, better risk control, and faster execution. Its real-time adaptability is notable, though retraining is needed for new markets. Ethical concerns about market fairness remain.
What it examines
This paper introduces TiMi, a multi-agent system for quantitative financial trading. TiMi uses large language models for semantic analysis, code programming, and mathematical reasoning, aiming to create rational, efficient trading bots that separate strategy development from real-time execution, improving profitability and risk control in dynamic markets.
What it concludes
TiMi shows strong, stable performance across various financial markets, outperforming existing methods in profitability and risk management. Its approach can be applied to automated trading, risk control, and strategy optimization. Limitations include adaptation to new markets, and future work may focus on customizable agentic trading systems and ethical considerations.
Evidence objects
TiMi, a multi-agent trading system powered by large language models, introduces a rationality-driven architecture that separates complex strategy design from rapid, minute-level execution for high efficiency and robust risk control.
key_findings bullet 1 · key_findings · validation V0
Using a two-tier approachmacro market analysis and micro pair-specific customizationplus layered programming and closed-loop mathematical optimization, TiMi adapts strategies in real time to seize short-term market opportunities missed by competitors.
key_findings bullet 2 · key_findings · validation V0
In live tests on 200+ stock and crypto pairs, TiMi outperformed traditional and LLM-based methods in annual returns, risk metrics, and latency, though retraining is needed for new markets and ethical concerns remain.
key_findings bullet 3 · key_findings · validation V0
This paper presents TiMi, a rationality-driven multi-agent system for quantitative finance, uniquely decoupling strategy development from minute-level deployment. Leveraging LLMs for semantic, coding, and mathematical tasks, its layered architecture and closed-loop optimization are empirically validated across 200+ trading pairs, offering novel, impactful advances for financial AI agents and algorithmic trading.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … Building upon the mechanical rationality, we formulate a multi-agent architecture leveraging specialized LLM capabilities in semantic analysis, code programming, and …
Source row: 2057 · abstract type: snippet