LO
CapabilityCognitive automation2025-05-21
An agent framework reached up to twice the annualised returns of classical factor libraries in backtests while using 70% fewer factors
Quantitative analystoccupation page →
Event date / reported
2025-05-21
Evidence stage
CapabilityA demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
Tasks this bears on
Signal research and strategy design
Finding patterns in data that predict returns or risk, and turning them into trading or investment strategies.
Being augmented✓ Evidence-backed
Where this applies
A research framework in which agents propose, implement and test factors and models. The abstract reports that it achieves up to 2X higher annualised returns than classical factor libraries using 70% fewer factors, and outperforms deep time-series models on real markets. The results are backtests, not live trading, and the authors' employer distributes the framework.
What this means
Agents can run the research loop — propose a factor, code it, test it — and beat simple baselines in backtests. Whether a backtest survives the market, and who answers for it, is still the quant's problem.
What it does not yet show
Backtests by the framework's developer; not live results.
What you can check
Open arXiv 2505.15155 and find "70% fewer factors" in the abstract.
Does it change the assessment?
No — and this stage does not move it either. A "Capability" record is real evidence, but it does not upgrade a task judgement on its own. The 1 linked judgement above stand where they were.
Source
Li, Yang, Yang et al. (Microsoft Research Asia) — R&D-Agent-Quant: A Multi-Agent Framework for Data-Centric Factors and Model Joint Optimization, arXiv 2505.15155 (submitted 21 May 2025) · verified 2026-09-29 · Claude (VOLO agent) · interpreted 2026-09-29 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.