Cognitive alpha discovery
Exploring wider and more interpretable spaces of investment hypotheses through reasoning, code generation, and evolutionary feedback.
Research
Research, tested beyond the paper.
From published research and ACL recognition to independent replication, end-to-end systems, and live-market validation.
Published research
PaperPaperPaper
Cognitive Alpha Mining via LLM-Driven Code-Based Evolution
A framework combining code-level alpha representation, LLM reasoning, and evolutionary search for automated and interpretable signal discovery.
XALPHA: A Memory-Driven AI Quant Researcher for Hypothesis-to-Code Alpha Discovery
A memory-driven AI quant researcher that turns alpha discovery into a continuous hypothesis-to-code research process.
MicroTSM: Financial Time-Series Models with Market Microstructure Features
A domain-specific financial time-series model showing how Level-2 and tick-level market microstructure data, model scale, and specialized architecture can improve forecasting and downstream trading performance.
Research process
From hypothesis to code, validation, and continuous refinement.
Agentic research is an iterative system: ideas become executable hypotheses, evidence challenges them, and each cycle informs the next.
Research directions
Questions that shape the system.
Agent systems for markets
Coordinating specialized agents across research, evidence review, portfolio reasoning, and continuous learning.
Validation and governance
Testing robustness, interpretability, model risk, and accountability before increasingly autonomous systems move closer to capital.