Annan Wang, Yuchen Xie, Yu Cen, Leon Liangyu Chen, Qianyu Julie Zhu, Xin Li, Tianyi Wei, Weisi Lin (Nanyang Technological University) · Flow-Matching Trading

AllocFlow

AllocFlow is an end-to-end flow-matching framework that generates joint, signed position weights across multiple assets directly from raw market histories, using a cross-asset flow Transformer. Training combines imitation of a teacher strategy with group-relative reinforcement learning on portfolio-level financial rewards. On hourly cryptocurrency perpetual futures, RL refinement raises the imitation policy's annualised net return from 27.79% to 51.38% and its Calmar ratio from 0.955 to 4.043, and the model reaches a Calmar ratio of 4.90 on six months of unseen 2026 data without retraining.

Open the source ↗

Similar Research & Safety releases

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.