LLM agents improve financial strategy development reliability with QuantHarness
Reliable Before Profitable: QuantHarness for Governing Financial Strategy Development with LLM Agents
Artificial Intelligence
Summary
Building financial strategies with AI can be tricky because small mistakes or hidden changes might make the strategy unreliable. The authors created QuantHarness, a system that carefully checks each step when an AI builds financial strategies, making sure the process is clear and trustworthy. They also made a benchmark called FinRelBench to test how well these systems handle making strategies reliably. QuantHarness performed better than other systems and helps keep track of decisions during development.
What this means in practice
- •For quantitative finance teams: Ensure trustworthy end-to-end development of algorithmic trading strategies with clear evidence and controlled intermediate steps.
- •For financial software developers: Build AI-driven tools that convert natural language requirements into verifiable and testable financial strategies.
Authors
Jiasheng Mao, Jiacheng Liu, Hengwei Ye, Zheng Tian
Abstract
LLM agents can turn natural-language financial research requests into executable and backtested strategies, but a successful run may still implement the wrong rule, use information unavailable at decision time, or obscure how repairs and candidate selection changed the strategy. Because these failures arise throughout development, final code and performance metrics alone are insufficient to establish either fidelity to the request or the validity of the development process. We therefore address the reliability of financial strategy development at the trajectory level: the system must control how intermediate outputs become downstream state and preserve the evidence needed to justify the terminal outcome. We introduce QuantHarness, a two-layer agent harness in which a finance-specific Workflow structures specification, data grounding, construction, testing, and review, while Harness Core enforces stage-specific information, action, and commitment boundaries. We also introduce FinRelBench, a benchmark that evaluates strategy-development reliability through execution success, requirement satisfaction, and runtime evidence. QuantHarness achieves the highest Task Success at 90.72\%, compared with 74.23\%--84.54\% for the comparison systems. Controlled stress tests show that its boundaries contain targeted violations, while structured run records improve evidence retrieval and support decision replay under the evaluated protocol.