arXiv subjects

Zhongmin Zhang

Publications and source records attributed to Zhongmin Zhang.

6 recordsLinked to original sources

arXiv2026

GALA: Generative Aligned Learning for Adaptive Multimodal Representation in the Taobao Shangou Recommender System

Modern recommender systems in food delivery increasingly leverage multimodal signals, including images, text, and user interaction histories, to enhance user experience, yet effective fusion of these heterogeneous modalities remains challenging, hindering both the joint modeling of multimodal signals and adaptation to evolving user intent. In mainstream two-stage approaches, the separation between content-semantic pretraining of image-text encoders and behavior-driven ranking models limits alignment between semantic understanding and user behavior patterns. To address these issues, we present GALA, a three-stage pipeline whose core innovation lies in an intermediate "generative RL alignment" stage that constructs multimodal pretraining data from user behavior and refines it via conversion-based rewards, effectively bridging the pretraining-fine-tuning gap to align with downstream objectives. GALA comprises three stages: first, behavior-aware triplet pretraining on query-image-text pairs from search logs to early capture user intent and content preferences; second, a novel intermediate stage that refines multimodal embeddings through reward-driven optimization (GRPO) to dynamically align them with user behavior and bridge the pretraining-fine-tuning gap; and finally, integration of multimodal and ID embeddings via adaptive gating with a hybrid loss, preserving multimodal contributions under long-term ID-dominant training. GALA has been deployed in the production environment at Taobao Shangou, serving over 200 million daily active users. Compared with state-of-the-art (SOTA) methods, it delivers consistent offline gains of +0.12/+0.20 AUC along with better PCOC metrics. Large-scale online A/B tests further report a 0.55 percent increase in order volume, confirming GALA's effectiveness at industrial scale and its robustness across diverse demand patterns.

cs.IR↗

arXiv2024

Single-molecule Automata: Harnessing Kinetic-Thermodynamic Discrepancy for Temporal Pattern Recognition

Molecular-scale computation is crucial for smart materials and nanoscale devices, yet creating single-molecule systems capable of complex computations remains challenging. We present a theoretical framework for a single-molecule computer that performs temporal pattern recognition and complex information processing. Our approach introduces the concept of an energy seascape, extending traditional energy landscapes by incorporating control parameter degrees of freedom. By engineering a kinetic-thermodynamic discrepancy in folding dynamics, we demonstrate that a linear polymer with $N$ binary-state foldable units can function as a deterministic finite automaton, processing $2^N$ configurations. The molecule's dominant configuration evolves deterministically in response to mechanical signals, enabling recognition of complex temporal patterns. This design allows complete state controllability through non-equilibrium driving protocols. Our model opens avenues for molecular-scale computation with applications in biosensing, smart drug delivery, and adaptive materials. We discuss potential experimental realizations using DNA nanotechnology. This work bridges the gap between information processing devices and stochastic molecular systems, paving the way for sophisticated molecular computers rivaling biological systems in complexity and adaptability.

cond-mat.stat-mech↗

arXiv2024

2D capsid formation within an oscillatory energy landscape: orderly self-assembly depends on the interplay between a dynamic potential and intrinsic relaxation times

Multiple dissipative self-assembly protocols designed to create novel structures or to reduce kinetic traps have recently emerged. Specifically, temporal oscillations of particle interactions have been shown effective at both aims, but investigations thus far have focused on systems of simple colloids or their binary mixtures. In this work, we expand our understanding of the effect of temporally oscillating interactions to a two-dimensional coarse-grained viral capsid-like model that undergoes a self-limited assembly. This model includes multiple intrinsic relaxation times due to the internal structure of the capsid subunits and, under certain interaction regimes, proceeds via a two-step nucleation mechanism. We find that oscillations much faster than the local intrinsic relaxation times can be described via a time averaged inter-particle potential across a wide range of interaction strengths, while oscillations much slower than these relaxation times result in structures that adapt to the attraction strength of the current half-cycle. Interestingly, oscillation periods similar to these relaxation times shift the interaction window over which orderly assembly occurs by enabling error correction during the half-cycles with weaker attractions. Our results provide fundamental insights to non-equilibrium self-assembly on temporally variant energy landscapes.

cond-mat.soft↗

arXiv2024

Stochastic Distinguishability of Markovian Trajectories

The ability to distinguish between stochastic systems based on their trajectories is crucial in thermodynamics, chemistry, and biophysics. The Kullback-Leibler (KL) divergence, $D_{\text{KL}}^{AB}(0,τ)$, quantifies the distinguishability between the two ensembles of length-$τ$ trajectories from Markov processes A and B. However, evaluating $D_{\text{KL}}^{AB}(0,τ)$ from histograms of trajectories faces sufficient sampling difficulties, and no theory explicitly reveals what dynamical features contribute to the distinguishability. This letter provides a general formula that decomposes $D_{\text{KL}}^{AB}(0,τ)$ in space and time for any Markov processes, arbitrarily far from equilibrium or steady state. It circumvents the sampling difficulty of evaluating $D_{\text{KL}}^{AB}(0,τ)$. Furthermore, it explicitly connects trajectory KL divergence with individual transition events and their waiting time statistics. The results provide insights into understanding distinguishability between Markov processes, leading to new theoretical frameworks for designing biological sensors and optimizing signal transduction.

cond-mat.stat-mech↗

arXiv2023

Non-equilibrium Theoretical Framework and Universal Design Principles of Oscillation-Driven Catalysis

At stationary environmental conditions, a catalyst's reaction rates may be restricted by thermodynamic laws, and certain performances can never be achieved (e.g., catalysts can not change the free energy difference between reactants and products). However, it has been reported that if environments change rapidly, catalysts can be driven away from stationary states and exhibit anomalous performance. We present a general geometric non-equilibrium theory to describe and explain anomalous catalytic behaviors in rapidly oscillating environments that exceed the steady-state restrictions. It leads to a universal design principle of novel catalysts with oscillation-pumped performances. Even though a catalyst at various environmental conditions cannot be described by a single free energy landscape, we propose a novel control-conjugate landscape to encode the reaction rates over a continuous range of control parameters $λ$, which is inspired by the exponential form of the Arrhenius law. The control-conjugate landscape significantly simplifies the design principle and makes it applicable to large-amplitude environmental oscillations. The design principle is demonstrated by two examples, (1) inverting a spontaneous reaction to synthesize high-free-energy molecules and (2) speeding up reactions without utilizing low activation barriers. In both examples, catalysts autonomously harness energy from non-equilibrium environments to enable such functionalities.

cond-mat.stat-mech↗

arXiv2022

Energy Landscape Design Principle for Optimal Energy Harnessing by Catalytic Molecular Machines

Under temperature oscillation, cyclic molecular machines such as catalysts and enzymes could harness energy from the oscillatory bath and use it to drive other processes. Using a novel geometrical approach, under fast temperature oscillation, we derive a general design principle for obtaining the optimal catalytic energy landscape that can harness energy from a temperature-oscillatory bath and use it to invert a spontaneous reaction. By driving the reaction against the spontaneous direction, the catalysts convert low free energy product molecules to high free energy reactant molecules. The design principle, derived for arbitrary cyclic catalysts, is expressed as a simple quadratic objective function that only depends on the reaction activation energies, and is independent of the temperature protocol. Since the reaction activation energies are directly accessible by experimental measurements, the objective function can be directly used to guide the search for optimal energy-harvesting catalysts.

cond-mat.stat-mech↗