AQuA: Recursively Self-Improving Quantitative Trading Research Agents
Researchers have developed AQuA, a system that uses language models to improve quantitative trading research. It consists of two separate systems: one for discovering symbolic factors and another for developing trainable models. Each system independently improves its own research process using validated evidence from earlier experiments. The factor system discovers factors that combine into a signal with an information coefficient of about 0.190, while the model system reache
Researchers have developed AQuA, a system that uses language models to improve quantitative trading research. It consists of two separate systems: one for discovering symbolic factors and another for developing trainable models. Each system independently improves its own research process using validated evidence from earlier experiments. The factor system discovers factors that combine into a signal with an information coefficient of about 0.190, while the model system reaches an information coefficient of +0.0843 on US equities. A strategy based on this system's output is positive in every year from 2021 to 2025.
---
Why it matters: This research matters because it demonstrates a potential approach to improving quantitative trading research through recursive self-improvement, which could lead to more effective investment strategies.
Source: https://arxiv.org/abs/2608.12841
This article was originally published at: https://arxiv.org/abs/2608.12841