TraderHarness
What it is
A leakage-resistant backtesting framework and training data synthesizer for LLM trading agents, specifically designed for A-share markets. It enforces strict temporal masking and entity anonymization to prevent data contamination, provides deterministic replay of agent actions, and exports full-fidelity trajectories for model training.
The problem it solves
Standard LLM backtesting suffers from data leakage (models recognizing historical dates/companies) and non-reproducibility, making it difficult to trust evaluation results or generate consistent training data.
What you could use it for
- Backtesting LLM trading agents on historical A-share data without future data leakage
- Generating high-fidelity training trajectories for reinforcement learning or SFT of trading models
- Comparing the performance of different LLM models or agent strategies in a controlled, reproducible environment
- Auditing and replaying agent decision-making processes for research and debugging
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.