Xinjie Chen, Minpeng Liao, Guoxin Chen, Chengxi Li, Biao Fu, Kai Fan, Xinggao Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Biopharmaceutical processes typically generate abundant high-frequency online sensor data but suffer from sparse and delayed offline quality measurements, creating a “data-rich but label-poor” dilemma that hinders effective process monitoring. Furthermore, these processes are subject to significant distribution shifts due to batch-to-batch variability and time-varying metabolic states. To address these challenges, this article proposes a novel online time-series contrastive learning framework for soft sensing, validated on an industrial erythromycin fermentation process. We introduce a mechanism-informed soft contrastive learning strategy that utilizes information derived from domain knowledge to construct instance-level and temporal similarity matrices, guiding the model to learn physically meaningful representations from unlabeled data. In addition, we develop a dual-timescale online learning architecture comprising a slow branch for robust representation learning and a fast branch for real-time adaptation. Experimental results on a large-scale industrial dataset demonstrate that the proposed framework significantly outperforms state-of-the-art time-series contrastive learning baselines, particularly in predicting complex rheological indicators like broth viscosity under temporal distribution shifts.

Time series modeling presents unique challenges due to autocorrelation in both historical data and future sequences. While current research predominantly addresses autocorrelation within historical data, the correlations among future labels are often overlooked. Specifically, modern forecasting models primarily adhere to the Direct Forecast (DF) paradigm, generating multi-step forecasts independently and disregarding label correlations over time. In this work, we demonstrate that the learning objective of DF is biased in the presence of label correlation. To address this issue, we propose the Frequency-enhanced Direct Forecast (FreDF), which mitigates label correlation by learning to forecast in the frequency domain, thereby reducing estimation bias. Our experiments show that FreDF significantly outperforms existing state-of-the-art methods and is compatible with a variety of forecast models. Code is available at https://github.com/Master-PLC/FreDF.

This study presents a nonlinear feedback regulator for three-dimensional overhead cranes that exploits composite outputs to deliver potent sway suppression. Dedicated barrier functions confine these composite signals within set limits. Owing to the simple structure, the control scheme maintains oscillation suppression under velocity and cable length uncertainties. We validate stability through a Lyapunov-based proof augmented by LaSalle’s invariance principle. Simulation results confirm that the controller achieves trolley positioning and effectively cancels oscillations under external disturbances.

This study proposes a nonlinear regulation controller for 3D overhead cranes, capable of achieving payload sway suppression under complex conditions. Utilizing a special form of barrier functions, constraints on the trolley motion and payload sway are derived. By analyzing the nonlinear terms, parameter estimation is incorporated to the control law, eliminating the need for prior knowledge of crane dynamics. To ensure smooth operation under varying transportation distances, a saturation function is employed to constrain the control torque generated by regulation errors. Within the Lyapunov framework, LaSalle’s invariance principle is invoked to demonstrate asymptotic convergence of the system states. Simulations validate the theoretical claims, including robustness under different transfer scenarios.

Traditional agriculture often relies heavily on pesticides for distributed pest management, which increases production costs and can lead to environmental pollution. The inefficiency and environmental impact of these methods highlight the need for more sustainable solutions. Artificial Intelligence Internet of Things (AIoT) enables precise monitoring and prediction of pest infestations in agriculture, facilitating targeted interventions to effectively reduce crop loss rates. This paper proposes a unified pest prevention and control system based on Agricultural Social Internet of Things (Agri-Social IoT) and artificial intelligence algorithm. By social relationships among devices, the system aims to enhance agricultural production efficiency through unified pest prediction and prevention. To achieve rapid and accurate detection of plant pests, this paper introduces an innovative detection method named Ghost-YOLO-ShuffleAttention (GhostYOLOSA). This artificial intelligence method improves the capture of spatial relationships and contextual information, enhancing the detection accuracy for small targets. It significantly reduces the computational load and model parameters while ensuring high detection accuracy, making it suitable for resource-limited devices. Experimental results from datasets and real-world applications demonstrate that the unified pest control system performs well in pest detection and prevention tasks, promoting sustainable agricultural development.

The industrial manufacturing industry requires user devices (UDs) to process massive data, leading to latency and energy consumption. To achieve low-latency and low-power task execution, we propose an industrial intelligent manufacturing system utilizing mobile edge computing. A dynamic computation offloading and resource allocation problem is formulated in multiserver multi-user scenarios to balance latency and energy cost. To solve this optimization problem, multi-agent deep reinforcement learning (MADRL) offers a theoretical framework. However, the widely used centralized training and decentralized execution (CTDE) scheme has two major drawbacks. First, it is unsuitable for scenarios without a central controller for global information collection. Second, it incurs significant communication overhead. Consequently, the decentralized training and execution (DTDE) scheme becomes necessary. However, DTDE introduces nonstationarity by treating other UDs as part of the environment, which leads to non-convergence. To address these issues, we design a distributed partial communication-based computation offloading and resource allocation algorithm (DPC-CORAA). This algorithm establishes a partial communication model based on the decentralized partially observable stochastic game (Dec-POSG) framework. It also incorporates the multi-agent deep deterministic policy gradient method under the DTDE scheme. The proposed method enables UDs to exchange information with neighbors to estimate the global decision, reducing communication cost. It also ensures theoretical convergence of this estimation to the real decision, serving as a local observation for independent strategy learning. Simulation results demonstrate that DPCCORAA achieves stable convergence, whereas the general DTDE scheme fails to converge. When contrasted with MADRL using CTDE, DPC-CORAA delivers superior performance in reducing latency and energy consumption, particularly in large-scale scenarios.

Multivariate soft sensors seek to provide accurate estimation of multiple quality variables through the analysis of measurable process variables, representing a significant advance over the traditional focus on single-quality variable sensors within industrial manufacturing. Current progress stays in applying parameter-sharing neural architectures while ignoring two fundamental issues: (1) catastrophic interference, where the indiscriminate sharing of parameters degrades performance due to the discrepancy of objectives; (2) seesaw optimization, where the optimizer overly focuses on one dominant yet simple objective at the expense of others. To address these issues, we reformulate multivariate soft sensors as a multi-objective optimization problem and propose the Task-aware Mixture-of-Experts framework for achieving the Pareto optimum (TMoE-P). Specifically, to handle issue(1), we propose an Objective-aware Mixture-of-Experts (OMoE) module, which consists of objective-specific and objective-shared experts to realize parameter sharing while accommodating the discrepancy between objectives. To handle issue(2), we devise a Pareto Objective Weighting (POW) module, which dynamically balances the weights of learning objectives to approximate the Pareto optimum among competing objectives. Our evaluations on a public soft sensor benchmark showcase TMoE-P’s superior performance, confirming its enhanced accuracy and robustness.Note to Practitioners—Addressing the burgeoning complexity of estimating multiple quality variables in industrial manufacturing processes, this study introduces a novel Task-aware Mixture-of-Experts framework aiming for the Pareto Optimum (TMoE-P). This framework mitigates the issues of catastrophic interference and seesaw optimization by achieving a delicate balance between parameter sharing and maintaining distinctness between objectives, guiding towards the Pareto optimum. The empirical findings affirm the framework’s capability to adeptly handle the complex interplay of variables, potentially enhancing the efficiency and reliability of industrial processes. While initially developed for multivariate data in industrial applications, the TMoE-P framework’s modular and adaptable nature facilitates its extension to a broad spectrum of data structures, optimizers, and neural architectures. Its adaptability offers practitioners a valuable tool to fulfill specific task requirements, making a modest contribution to industrial process control and monitoring.

Multivariate soft sensors are critical in industrial manufacturing for providing precise estimations of multiple quality variables and ensuring data reliability and completeness. Existing methods predominantly focus on parameter-sharing architectures while overlooking two fundamental issues: 1) catastrophic interference, where sharing parameters across all tasks leads to performance degradation; 2) uncontrollable optimization, where the optimizer lacks controllability over task priorities, misleading specific tasks converge to unexpected values. To handle these issues and enhance multivariate modeling, we propose a controllable Mixture-of-Experts (ControlMoE) framework for industrial soft sensors, consisting of a Mixture-of-Sequential-Experts (MoSE) module and a Proportional-Integral-Derivative Calibrating (PIDC) module. The MoSE module integrates sequential expert networks with task-specific gating networks, enabling parameter sharing while preserving task distinctions to mitigate catastrophic interference. The PIDC module employs an innovative PID controller to dynamically calibrate task learning weights, ensuring expected outcomes through feedback control and enhancing optimization controllability. Evaluations on two industrial datasets from Chinese nuclear monitoring stations demonstrate that ControlMoE achieves superior accuracy and controllability in multivariate soft sensor modeling. Note to Practitioners—This paper tackles the challenges in applying multivariate soft sensors within industrial manufacturing processes, particularly in environments such as nuclear monitoring where precision and control over multiple quality variables are critical. Traditional methods often suffer from performance degradation due to parameter sharing across all tasks-known as catastrophic interference–and lack control in optimizing multiple tasks simultaneously, leading to suboptimal outcomes. Our proposed framework, ControlMoE, introduces a novel architecture that combines sequential expert networks with a dynamic weight adjustment mechanism using PID controllers. This enables precise control over task-specific optimizations, ensuring that the desired importance and sequence of tasks are maintained, thereby enhancing both accuracy and operational control. We believe this framework could be widely applicable in various industrial settings where multivariate monitoring and control are necessary, improving not only performance but also the reliability of the data produced for critical decision-making.

Accurate imputation of missing data is pivotal in real-world industrial applications. Traditional direct imputers, which utilize basic statistics to replace missing elements, offer a practical solution but struggle to adapt to the complex patterns in industrial data, leaving a gap in the research landscape. This study explores the untapped potential of direct imputers, enhancing their adaptability and capacity to handle complex patterns in industrial data through optimal transport (OT) theory, with a focus on preserving local sample-wise similarity as an exemplar. To these ends, we construct a Local Similarity Preserved Transport (LSPT) problem, with a solution algorithm based on the Frank-Wolfe technique to compute transport cost. Subsequently, we propose the LSPT-D framework, which employs the transport cost of LSPT for distribution matching, directing the gradient flow to the missing data points to update the imputations directly. This strategy maintains local similarity throughout the imputation process thereby enhancing the overall imputation quality. Our experiments demonstrate that LSPT-D outperform various baselines in industrial missing data imputation. Note to Practitioners—Accurate missing data imputation is essential for enhancing the reliability of data analytics and reducing decision-making risks in industrial automation. This study introduces LSPT-D, a non-parametric imputation technique based on OT technology. Unique to LSPT-D is its ability to preserve local similarity during the imputation process, rendering it particularly advantageous for datasets with varying operational phases and load conditions. In industrial applications, LSPT-D not only significantly improves imputation quality compared to various baseline methods but also maintains modest running costs. Additionally, it serves as an exemplar for developing OT-based imputation strategies that capitalize on the inherent properties of data to improve imputation performance. However, LSPT-D operates under the independent and identically distributed assumption and is thus best applied in scenarios where temporal dependencies, such as trends and seasonality, are minimal or have been previously neutralized.

Post-click conversion rate (CVR) estimation is a fundamental task in developing effective recommender systems, yet it faces challenges from data sparsity and sample selection bias. To handle both challenges, the entire space multitask models are employed to decompose the user behavior track into a sequence of exposure$\rightarrow $click$\rightarrow $conversion, constructing surrogate learning tasks for CVR estimation. However, these methods suffer from two significant defects: (1) intrinsic estimation bias (IEB), where the CVR estimates are higher than the actual values; (2) false independence prior (FIP), where the causal relationship between clicks and subsequent conversions is potentially overlooked. To overcome these limitations, we develop a model-agnostic framework, namely Entire Space Counterfactual Multitask Model (ESCM2), which incorporates a counterfactual risk minimizer within the entire space multitask framework to regularize CVR estimation. Experiments conducted on large-scale industrial recommendation datasets and an online industrial recommendation service demonstrate that ESCM2 effectively mitigates IEB and FIP defects and substantially enhances recommendation performance.

To achieve high-quality translation with low latency, a Simultaneous Speech Translation (SimulST) system relies on a policy module to decide whether to translate immediately or wait for additional streaming input, along with a translation model capable of effectively handling partial speech input. Prior research has tackled these components separately, either using ``wait-k'' policies based on fixed-length segments or detected word boundaries, or dynamic policies based on different strategies (e.g., meaningful units), while employing offline models for prefix-to-prefix translation. In this paper, we propose Divergence-Guided Simultaneous Speech Translation (DiG-SST), a tightly integrated approach focusing on both translation quality and latency for streaming input. Specifically, we introduce a simple yet effective prefix-based strategy for training translation models with partial speech input, and develop an adaptive policy that makes read/write decisions for the translation model based on the expected divergence in translation distributions resulting from future input. Our experiments on multiple translation directions of the MuST-C benchmark demonstrate that our approach achieves a better trade-off between translation quality and latency compared to existing methods.

AI4Biopharmaceutical is an important field of interest in both academia and industry. Due to the characteristics of biopharmaceutical industry data and the hardware limitation of real factories, biopharmaceutical indicators prediction models need to balance computational complexity and prediction effectiveness. We propose a low computational complexity hybrid model (LCCH) combing the biopharmaceutical mechanism with artificial intelligence for biopharmaceutical indicators prediction. We use the Doolittle method to reduce the computational complexity of the online decomposition phase to O(1), while employing the half-router attention incorporating biopharmaceutical mechanism to significantly reduce the computational complexity of the prediction phase. The model achieves the state-of-the-art results for the prediction of five biopharmaceutical indicators: amino nitrogen, reducing sugar, total sugar, bacterial concentration,and viscosity in the erythromycin pharmaceutical scenario. Compared to the other outstanding baseline models from the last three years, LCCH demonstrates the superiority and potential of the hybrid model for biopharmaceutical industrial applications.

This paper proposes a nonlinear feedback control for overhead cranes that offer satisfactory performance by taking advantages of only a composite output. Particularly, the construction of a barrier function keeps the composite output between predefined boundary values, which can enhance the safety of the system. Nonetheless, the controller with simple structure ensures the stabilization of the payload despite the presence of parametric uncertainties. To substantiate the stability proof, two analytical methodologies are employed: the Lyapunov technique and LaSalle’s invariance principle. The simulation evidences efficient positioning and oscillation elimination of the controller for various uncertain parameters, large initial errors and external disturbances, without tuning the gains of each term.

The increased frequency of forest fires in recent years has raised concerns about the high cost associated with traditional forest fire prevention methods. To address this issue, this paper presents a novel forest fire detection and segmentation application for UAV-assisted mobile edge computing system. Traditional target detection and segmentation systems for forest fire detection are often large and unsuitable for deployment on edge equipment such as UAVs. Deploying such models on edge gateways can also lead to high costs and delays. To overcome these challenges, this paper proposes a lightweight fire target detection and precision segmentation model that can be used on UAVs and other edge equipment. The proposed algorithm achieves more accurate image segmentation, thereby improving the efficiency of fire location. Additionally, an edge computing system is built to link the feedback of the edge model with the edge gateway, administrators, and other intelligent devices promptly. Extensive experiments with large datasets and in real environments demonstrate the efficacy of the proposed algorithm, effectively enhancing the efficiency of forest inspection and forest fire warning capabilities.

Specific emitter identification (SEI) uses the electromagnetic pulse signal sent by emitter to determine the emitter individual. In the actual complex electromagnetic environment, due to the interference of external signals and hardware failures, it is difficult to obtain sufficient and complete transmitter signal data. The missing data imputation methods are used to impute the emitter signal data. However, the existing imputation methods need to rely on the complete signal data to train the deep learning model, and the imputation error is large due to the long sequence characteristics of the signal. Therefore, a new specific emitter missing data imputation model is proposed, which is called bidirectional stackable recurrent generative adversarial imputation network (BiSRGAIN) including a generator and a discriminator. Specifically, the bidirectional stackable recurrent (BiSR) unit is designed to be used in generators and discriminators, which simplifies the traditional recurrent neural network (RNN) structure and improves parameter utilization and inference efficiency. The novel loss function can make the training of the model independent of the true value of the missing components, so the model can be trained in incomplete data. Extensive experiments are conducted on real-world dataset. The results show that the proposed model has lower errors under the scenario of high missing rate. In addition, the proposed model has higher parameter utilization and computational efficiency. Moreover, the completed signal data after imputation is used to identify specific emitters, and the results show that the data obtained by BiSRGAIN can achieve higher recognition accuracy.

Forecasting governing key factors in industrial bioprocesses is crucial for ensuring stability and efficiency in production. However, the accurate prediction is challenged by the strong coupling and uncertainty characteristic of industrial bioprocess data. To capture the common dynamics and comprehensively model the interrelationships among multivariate time series in bioprocesses, this study introduces a predictive model called the causality enhanced global-local graph neural network. A global-local decomposition module is first constructed utilizing time regularization, thereby, explicitly obtaining global and local bioprocess series while preserving temporal structure. Subsequently, we construct node embedding for both the global and local series. Finally, we presents an innovative graph generation module that creates an explicit causality graph based on transfer entropy and an implicit static-dynamic graph for the downstream graph neural network, considering causal information, static and dynamic dependencies among variables. Application results based on real industrial bioprocess data demonstrate that this method has high predictive accuracy.

Missing data imputation is a critical aspect of the Industrial Internet-of-Things (IIoT), which is uniquely challenged by local relationships within data due to different operational contexts and phases. Current imputation methods struggle to accommodate local relationships due to their black-box nature or limited capacity. To bridge this gap, we approach data imputation as a distribution alignment problem and leverage optimal transport to instantiate it for enhanced capacity. Specifically, we first introduce the similarity preserved optimal transport (SPOT) problem, with a conditional gradient solution to compute the transport cost. Subsequently, we propose the SPOT for imputation (SPOT-I) framework. It minimizes the transport cost of SPOT for distribution alignment and uses the gradient to update imputations, which maintains local similarity and refines imputation due to the characteristics of SPOT. Experiments on IIoT datasets showcase the superiority of SPOT-I over state-of-the-art imputation methods.

Monitoring of energy conversion process confronts great difficulties due to extreme value jumps or data packet loss under extreme operating conditions, consequently resulting in data missing. To tackle these issues, researchers propose diffusion-based time series imputation approaches. However, there are critical limitations of methods adopted by these models: first, dense reverse inference (DRI), where conventional diffusion methods suffer from time-consuming imputation procedures and thus the loss of effective information over a long transmission process; second, indirect prediction strategy (IPS), which causes inaccurate temporal data imputation results. To address these limitations, we propose denoising diffusion straightforward models (DDSMs) for missing data imputation in energy conversion process monitoring. Specifically, based on the conditional mechanism, we innovatively use the accelerated sampling strategy in temporal models to reduce the lengthy inference time caused by DRI and propose a well-designed straightforward training algorithm for a straightforward estimation to improve the imprecise inference results acquired by IPS. Extensive experiments on real-world dataset demonstrate the effectiveness and superiority of DDSM and our approach consistently outperforms previous temporal generative methods significantly.

Accurate estimation of multiple quality variables is critical for building industrial soft sensor models, which have long been confronted with data efficiency and negative transfer issues. Methods sharing backbone parameters among tasks address the data efficiency issue; however, they still fail to mitigate the negative transfer problem. To address this issue, a balanced mixture-of-experts (BMoE) is proposed in this work, which consists of a multigate mixture-of-experts module and a task gradient balancing (TGB) module. The mixture-of-experts module aims to portray task relationships, while the TGB module balances the gradients among tasks dynamically. Both of them cooperate to mitigate the negative transfer problem. Experiments on the typical sulfur recovery unit demonstrate that BMoE models task relationship and balances the training process effectively, and achieves better performance than baseline models significantly.

Triplet loss with batch hard mining (TriHard loss) is an important variation of triplet loss inspired by the idea that hard triplets improve the performance of metric leaning networks. However, there is a dilemma in the training process. The hard negative samples contain various quite similar characteristics compared with anchors and positive samples in a batch. Features of these characteristics should be clustered between anchors and positive samples while are also utilized to repel between anchors and hard negative samples. It is harmful for learning mutual features within classes. Several methods to alleviate the dilemma are designed and tested. In the meanwhile, an element-weighted TriHard loss is emphatically proposed to enlarge the distance between partial elements of feature vectors selectively which represent the different characteristics between anchors and hard negative samples. Extensive evaluations are conducted on Market1501 and MSMT17 datasets and the results achieve state-of-the-art on public baselines.