Research, models, and systems for the next generation of financial AI
Freeride combines frontier research, proprietary financial models, and trading systems to advance AI-led decision making.
STAR Analyst: Self-Tuning Alpha Research
William F. Shen, Alex Iacob, Zichen Zhang, Daoheng Wang, Wenyi Wang, Rui Liang, Yulong Zhang, Xinchi Qiu, Nicholas D. Lane · University of Cambridge, Freeride AI
STAR shifts the object of evolution from individual alpha formulas to the alpha researcher itself. It couples an LLM-based researcher with a metacognitive module, while fixed validation and test boundaries preserve evaluation integrity. Under a leakage-controlled forward protocol, evolved STAR researchers produced substantially stronger alphas than the base model.
One stack across research, models, and live systems
Research, model training, and product feedback evolve together.
How specialized agents reason, disagree, and converge on actionable views.
Domain-specific models trained on proprietary data, workflows, and financial reasoning tasks.
Research and models are refined through system behavior, operating data, and live decision feedback.
What STAR demonstrates
Research findings, not product return claims.
Generated-alpha median IR
Evolved STAR researchers shifted the held-out generated-alpha IR distribution upward in the reported CSI300 2026Q1 forward protocol.
Reported quarterly excess return
The paper reports 17% quarterly excess return, IR of 1.54, and maximum drawdown below 3% in downstream evaluation.
Operator vocabulary expansion
Qualitative analysis suggests the gains came from genuine researcher evolution, not only selecting better factor nodes.
Daily brief and highlighted names
Start with the latest market brief and the day’s recommended names, then open the full reports when needed.
What the market is highlighting today
Loading the latest morning brief and top picks...