Author name cluster

Zeyuan Wang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

9 papers

2 author rows

AAAI Conference 2026 Conference Paper

One-Step Generative Policies with Q-Learning: A Reformulation of MeanFlow

Zeyuan Wang
Da Li
Yulin Chen
Ye Shi
Liang Bai
Tianyuan Yu
Yanwei Fu

We introduce a one-step generative policy for offline reinforcement learning that maps *noise* directly to *actions* via a *residual reformulation* of MeanFlow, making it compatible with Q-learning. While one-step Gaussian policies enable fast inference, they struggle to capture complex, multimodal action distributions. Existing flow-based methods improve expressivity but typically rely on distillation and two-stage training when trained with Q-learning. To overcome these limitations, we propose to reformulate MeanFlow to enable *direct noise-to-action generation* by integrating the velocity field and noise-to-action transformation into a single policy network—eliminating the need for separate velocity estimation. We explore several reformulation variants and identify an effective *residual formulation* that supports expressive and stable policy learning. Our method offers three key advantages: 1) efficient one-step noise-to-action generation, 2) expressive modelling of multimodal action distributions, and 3) efficient and stable policy learning via Q-learning in a single-stage training setup. Extensive experiments on 73 tasks across the OGBench and D4RL benchmarks demonstrate that our method achieves strong performance in both offline and offline-to-online reinforcement learning settings.

PDF Details DOI

ICLR Conference 2025 Conference Paper

COMBO: Compositional World Models for Embodied Multi-Agent Cooperation

Hongxin Zhang
Zeyuan Wang
Qiushi Lyu
Zheyuan Zhang
Sunli Chen
Tianmin Shu
Behzad Dariush
Kwonjoon Lee

In this paper, we investigate the problem of embodied multi-agent cooperation, where decentralized agents must cooperate given only egocentric views of the world. To effectively plan in this setting, in contrast to learning world dynamics in a single-agent scenario, we must simulate world dynamics conditioned on an arbitrary number of agents' actions given only partial egocentric visual observations of the world. To address this issue of partial observability, we first train generative models to estimate the overall world state given partial egocentric observations. To enable accurate simulation of multiple sets of actions on this world state, we then propose to learn a compositional world model for multi-agent cooperation by factorizing the naturally composable joint actions of multiple agents and compositionally generating the video conditioned on the world state. By leveraging this compositional world model, in combination with Vision Language Models to infer the actions of other agents, we can use a tree search procedure to integrate these modules and facilitate online cooperative planning. We evaluate our methods on three challenging benchmarks with 2-4 agents. The results show our compositional world model is effective and the framework enables the embodied agents to cooperate efficiently with different agents across various tasks and an arbitrary number of agents, showing the promising future of our proposed methods. More videos can be found at https://umass-embodied-agi.github.io/COMBO

Details

NeurIPS Conference 2024 Conference Paper

DePLM: Denoising Protein Language Models for Property Optimization

Zeyuan Wang
Keyan Ding
Ming Qin
Xiaotong Li
Xiang Zhuang
Yu Zhao
Jianhua Yao
Qiang Zhang

Protein optimization is a fundamental biological task aimed at enhancing theperformance of proteins by modifying their sequences. Computational methodsprimarily rely on evolutionary information (EI) encoded by protein languagemodels (PLMs) to predict fitness landscape for optimization. However, thesemethods suffer from a few limitations. (1) Evolutionary processes involve thesimultaneous consideration of multiple functional properties, often overshadowingthe specific property of interest. (2) Measurements of these properties tend to betailored to experimental conditions, leading to reduced generalizability of trainedmodels to novel proteins. To address these limitations, we introduce DenoisingProtein Language Models (DePLM), a novel approach that refines the evolutionaryinformation embodied in PLMs for improved protein optimization. Specifically, weconceptualize EI as comprising both property-relevant and irrelevant information, with the latter acting as “noise” for the optimization task at hand. Our approachinvolves denoising this EI in PLMs through a diffusion process conducted in therank space of property values, thereby enhancing model generalization and ensuringdataset-agnostic learning. Extensive experimental results have demonstrated thatDePLM not only surpasses the state-of-the-art in mutation effect prediction butalso exhibits strong generalization capabilities for novel proteins.

PDF Details DOI

ICML Conference 2024 Conference Paper

Knowledge-aware Reinforced Language Models for Protein Directed Evolution

Yuhao Wang
Qiang Zhang 0026
Ming Qin
Xiang Zhuang
Xiaotong Li
Zhichen Gong
Zeyuan Wang
Yu Zhao 0009

Directed evolution, a cornerstone of protein optimization, is to harness natural mutational processes to enhance protein functionality. Existing Machine Learning-assisted Directed Evolution (MLDE) methodologies typically rely on data-driven strategies and often overlook the profound domain knowledge in biochemical fields. In this paper, we introduce a novel Knowledge-aware Reinforced Language Model (KnowRLM) for MLDE. An Amino Acid Knowledge Graph (AAKG) is constructed to represent the intricate biochemical relationships among amino acids. We further propose a Protein Language Model (PLM)-based policy network that iteratively samples mutants through preferential random walks on the AAKG using a dynamic sliding window mechanism. The novel mutants are actively sampled to fine-tune a fitness predictor as the reward model, providing feedback to the knowledge-aware policy. Finally, we optimize the whole system in an active learning approach that mimics biological settings in practice. KnowRLM stands out for its ability to utilize contextual amino acid information from knowledge graphs, thus attaining advantages from both statistical patterns of protein sequences and biochemical properties of amino acids. Extensive experiments demonstrate the superior performance of KnowRLM in more efficiently identifying high-fitness mutants compared to existing methods.

Details

ECAI Conference 2023 Conference Paper

Active Finetuning Protein Language Model: A Budget-Friendly Method for Directed Evolution

Ming Qin
Keyan Ding
Bin Wu 0025
Zhenping Li
Haihong Yang
Zeyuan Wang
Hongbin Ye
Haoran Yu

Directed evolution is a widely-used strategy of protein engineering to improve protein function via mimicking natural mutation and selection. Machine learning-assisted directed evolution (MLDE) approaches aim to learn a fitness predictor, thereby efficiently searching for optimal mutants within the vast combinatorial mutation space. Since annotating mutants is both costly and labor-intensive, how to efficiently sample and utilize informative protein mutants to train the predictor is a critical problem in MLDE. Previous MLDE works just simply utilized pre-trained protein language models (PPLMs) for sampling without tailoring to the specific target protein of interest, which has not fully exploited the potential of PPLMs. In this work, we propose a novel method, the Actively-Finetuned Protein language model for Directed Evolution(AFP-DE), which leverages PPLMs to actively sample and fine-tune themselves, continuously improving the model’s sampling and overall performance through iterations, to achieve efficient directed protein evolution. Extensive experiments have shown the effectiveness of our method in generating optimal mutants with minimal annotation effort, outperforming previous works even with fewer annotated mutants, making it budget-friendly for biological experiments.

Details

EAAI Journal 2023 Journal Article

Modified TODIM method based on cumulative prospect theory with Type-2 neutrosophic number for green supplier selection

Zeyuan Wang
Qiang Cai
Guiwu Wei

Today’s consumers’ awareness of environmental protection and social morality is forcing companies to implement sustainable supply chains. Choosing a green supplier can not only determine the sustainable development of the supply chain, but also protect the environment and enhance the company’s reputation, in line with the background of green development. Green supplier selection is a multi-attribute decision-making (MADM) problem. Type-2 neutrosophic number (T2NN) can be compatible with more fuzzy information, and the green supplier selection problem also has a lot of fuzzy information, which makes T2NN very suitable for supplier selection. In this paper, a new method named CPT-TODIM-CDMT is proposed for green supplier selection of T2NN set based on TODIM. A new distance measure named cosine distance measure is proposed for T2NN set, and is applied in TODIM method. In this new method, attribute weight is determined by entropy method and optimized by CPT. The gains and losses preference in the CPT is also applied to the step of calculating the dominance degree in the TODIM method. A case study is used to verify the method. Through the comparison of the results of different decision-making models and the sensitivity research of the coefficients in the model, the correctness, validity and stability of the model are verified.

Details DOI

ICLR Conference 2023 Conference Paper

Multi-level Protein Structure Pre-training via Prompt Learning

Zeyuan Wang
Qiang Zhang 0026
Shuangwei Hu
Haoran Yu
Xurui Jin
Zhichen Gong
Huajun Chen

A protein can focus on different structure levels to implement its functions. Each structure has its own merit and driving forces in describing some specific characteristics, and they cannot replace each other. Most existing function prediction methods take the tertiary structure as input, unintentionally ignoring the other levels of protein structures. Considering protein sequences can determine multi-level structures, in this paper, we aim to realize the comprehensive potential of protein sequences for function prediction. Specifically, we propose a new prompt-guided multi-task pre-training and fine-tuning framework, and the resulting protein model is called PromptProtein. Through the prompt-guided multi-task pre-training, we learn multiple prompt signals to steer the model to focus on different structure levels. We also design a prompt fine-tuning module to provide downstream tasks the on-demand flexibility of utilizing respective levels of structure information. Extensive experiments on function prediction and protein engineering show that PromptProtein outperforms state-of-the-art methods by large margins.

Details

AIIM Journal 2021 Journal Article

A novel method for clinical risk prediction with low-quality data

Zeyuan Wang
Josiah Poon
Shuze Wang
Shiding Sun
Simon Poon

In real-world data, predictive models for clinical risks (such as adverse drug reactions, hospital readmission, and chronic disease onset) are constantly struggling with low-quality issues, namely redundant and highly correlated features, extreme category imbalances, and most importantly, a large number of missing values. In most existing work, each patient is represented as a value vector with the fixed-length from some feature space, and missing values are forced to be imputed, which introduces much noise for prediction if the data set is highly incomplete. Besides, other challenges are either remaining unresolved or only partially solved when modeling, but without a systematic approach. In this paper, we propose a novel framework to address these low-quality problems, that we first treat patients as bags with the various number of feature-value pairs, called instances, and map them to an embedding space through our proposed feature embedding method to learn from it directly. In this way, predictive models can avoid the negative impact of missing data naturally. A novel multi-instance neural network is then connected, using two computational modules to deal with the problems of correlated and redundant features: multi-head attention and attention-based multi-instance pooling. They are capable of capturing the instance correlations and locating valuable information in each instance or bag. The feature embedding and multi-instance neural network are parameterized and optimized jointly in an end-to-end manner. Moreover, the training process is under both main and auxiliary supervision with focal loss functions to avoid the caveat of a highly imbalanced label set. This proposed framework is named AMI-Net3. We evaluate it on three suitable data sets from real-world settings with different clinical risk prediction tasks: adverse drug reaction of risperidone, schizophrenia relapse, and invasive fungi infection, respectively. The comprehensive experimental results demonstrate the effectiveness and superiority of our proposed method over competitive baselines.

Details DOI

IJCAI Conference 2021 Conference Paper

Reinforcement Learning Based Sparse Black-box Adversarial Attack on Video Recognition Models

Zeyuan Wang
Chaofeng Sha
Su Yang

We explore the black-box adversarial attack on video recognition models. Attacks are only performed on selected key regions and key frames to reduce the high computation cost of searching adversarial perturbations on a video due to its high dimensionality. To select key frames, one way is to use heuristic algorithms to evaluate the importance of each frame and choose the essential ones. However, it is time inefficient on sorting and searching. In order to speed up the attack process, we propose a reinforcement learning based frame selection strategy. Specifically, the agent explores the difference between the original class and the target class of videos to make selection decisions. It receives rewards from threat models which indicate the quality of the decisions. Besides, we also use saliency detection to select key regions and only estimate the sign of gradient instead of the gradient itself in zeroth order optimization to further boost the attack process. We can use the trained model directly in the untargeted attack or with little fine-tune in the targeted attack, which saves computation time. A range of empirical results on real datasets demonstrate the effectiveness and efficiency of the proposed method.

PDF Details DOI