Searcharxiv⌕ Search

arXiv subjects

Mingbo Zhou

Publications and source records attributed to Mingbo Zhou.

2 recordsLinked to original sources

A level-set method for shape and topology optimization of Stokes--Darcy flows with Beavers--Joseph--Saffman coupling

We propose a level-set method for shape and topology optimization of Stokes--Darcy flows with Beavers--Joseph--Saffman interface coupling. We formulate the Stokes--Darcy system with BJS interface conditions using a level-set representation and establish existence and uniqueness of its solution. We further prove existence of minimizers for the viscous-energy minimization and least-squares inverse problems within this framework. Continuous Fréchet differentiability of the control-to-state map is proved, and objective derivatives are derived using an adjoint approach. A reaction--diffusion evolution scheme combined with mixed finite elements is developed to find the optimized interface and accommodate topology changes. Two- and three-dimensional numerical experiments verify the derived sensitivities and demonstrate the effectiveness of the method in minimizing viscous energy dissipation and reconstructing interfaces from partial and noisy velocity observations.

math.OC↗

Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges

Reinforcement Learning from Human Feedback (RLHF) and related alignment paradigms have become central to steering large language models (LLMs) and multimodal large language models (MLLMs) toward human-preferred behaviors. However, these approaches introduce a systemic vulnerability: reward hacking, where models exploit imperfections in learned reward signals to maximize proxy objectives without fulfilling true task intent. As models scale and optimization intensifies, such exploitation manifests as verbosity bias, sycophancy, hallucinated justification, benchmark overfitting, and, in multimodal settings, perception--reasoning decoupling and evaluator manipulation. Recent evidence further suggests that seemingly benign shortcut behaviors can generalize into broader forms of misalignment, including deception and strategic gaming of oversight mechanisms. In this survey, we propose the Proxy Compression Hypothesis (PCH) as a unifying framework for understanding reward hacking. We formalize reward hacking as an emergent consequence of optimizing expressive policies against compressed reward representations of high-dimensional human objectives. Under this view, reward hacking arises from the interaction of objective compression, optimization amplification, and evaluator--policy co-adaptation. This perspective unifies empirical phenomena across RLHF, RLAIF, and RLVR regimes, and explains how local shortcut learning can generalize into broader forms of misalignment, including deception and strategic manipulation of oversight mechanisms. We further organize detection and mitigation strategies according to how they intervene on compression, amplification, or co-adaptation dynamics. By framing reward hacking as a structural instability of proxy-based alignment under scale, we highlight open challenges in scalable oversight, multimodal grounding, and agentic autonomy.

cs.LG↗