Mitigating Bias in Low-SNR Financial Reinforcement Learning via Quantum Representations

Jie Zhang, Song Guo, Zeyu Liu, Jingcai Guo, Xiaoyi Pang, Hualei Zhang, Xuanzhi Feng, Sing Lai, Kwong, Yuanchen Gao

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 8 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

Repositories linked to this paper

• ZeyuLIU-UST/FPQC-SAC-main

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The financial market is a typical low signal-to-noise ratio (SNR) setting, which often destabilizes off-policy maximum-entropy methods like Soft Actor-Critic (SAC). Specifically, noisy state representations may produce unreliable Q-value estimates, and bootstrapping amplifies these errors, forming a failure mode we call the "Financial Entropy Trap". In this paper, we propose FPQC-SAC, an efficient and plugand-play SAC variant that places a compact and bounded Parameterized Quantum Circuit (PQC) before the actor and critic networks to constrain feature propagation at the representation level, rather than filtering raw inputs or regularizing Q-values after bootstrapping. Notably, FPQC-SAC reduces the impact of extreme market fluctuations on Bellman target estimation, while trainable quantum entanglement preserves flexible cross-asset interactions. Empirical evaluations on real-world portfolio management tasks demonstrate that FPQC-SAC substantially enhances out-of-sample stability and cumulative returns by achieving a 66.89% relative gain in cumulative return over standard unconstrained SAC and outperforms the best continuous-control deep reinforcement learning baseline by ∼27%. open-source code is available at https://github.com/ZeyuLIU-UST/FPQC-SAC-main. At the core of this challenge lies sequential asset allocation: continuously determining when and how much to buy, hold, or sell. While traditional statistical models struggle to characterize complex marketstate transitions, Deep Reinforcement Learning (DRL) provides a promising end-to-end framework [27,28,21,47,38]. Among these, maximum-entropy reinforcement learning, represented by Soft Actor-Critic (SAC) [15], explicitly maximizes policy entropy to encourage stochastic exploration, aiming to maintain trading-policy diversity and generate high-quality decisions. However, the Actor-Critic architecture becomes significantly fragile in noise-dominant markets. Entropy-driven stochastic exploration does not necessarily gather effective information; instead, it may amplify noise-dominated update directions. This prevents the policy from stably capturing * Equal contribution.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2606.10448")
get_code_for_paper("2606.10448")
have("2606.10448")

Connect an agent — have() is free.