Reproduce “Hindsight Preference Optimization for Financial Time Series Advisory” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yanwei Cui, Guanghui Wang, Xing Zhang, Peiyang He, Ziyuan Li, Bing Zhu, Wei Qiu, Xusheng Wang and 2 more
The authors introduce Hindsight DPO, which uses realized outcomes to rank candidate financial advisories and create preferred and dispreferred examples for training, without exposing the forecaster to future data. On five S&P 500 stocks, a Qwen3-VL-4B model improves over the Qwen3-VL-235B teacher on directional and top-1 scenario accuracy in a held-out 2017 test, whose scope limits generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.