Where else to look
Can agents share what they find while pursuing genuinely different ideas?
Chutian Chen (Curtis)
Can an agent tell when it’s wrong?
I’m Curtis Chen, Co-founder and CEO of Quanta Edge and a Ph.D. candidate at HKUST. I research deep learning and AI agents in finance.
I study how agents test their ideas and change course when they fail. A price prediction puts an agent’s judgment on record so we can check it against what happens next.
Research questions
What should an agent learn from an experiment?
How much to trust a result
Does a finding hold up on fresh evidence? Does the agent’s confidence reflect how often it gets things right?
What to change next
Can an agent diagnose a failure and use it to design a better experiment?
Selected publication
EvoQuantLab
EvoQuantLab: Training-Free Dual-Loop Evolution for Auditable Quantitative Research
In this paper, my co-authors and I explore how an agent can use the results of a quantitative search to improve the search procedure itself.
Background
My path to AI for finance
Earlier work
I founded Futuristic Group during my Ph.D. and led its strategy research. My earlier work included recommendation algorithms at Alibaba, reinforcement learning for FX at Graphen and out-of-distribution robustness at Megvii.
Full experienceMy academic path
I’m a Ph.D. candidate at HKUST, researching AI for finance. I studied computer science at Columbia, engineering physics at Tsinghua, and economics at Peking University.
Download my CVGet in touch
I’d like to hear from people working on research agents, independent evaluation or learning from failed experiments.