Yize Chen

Yize Chen, Yushi Tan, Deepjyoti Deka

Recent advances in Machine Learning(ML) have led to its broad adoption in a series of power system applications, ranging from meter data analytics, renewable/load/price forecasting to grid security assessment. Although these data-driven methods yield state-of-the-art performances in many tasks, the robustness and security of applying such algorithms in modern power grids have not been discussed. In this paper, we attempt to address the issues regarding the security of ML applications in power systems. We first show that most of the current ML algorithms proposed in power systems are vulnerable to adversarial examples, which are maliciously crafted input data. We then adopt and extend a simple yet efficient algorithm for finding subtle perturbations, which could be used for generating adversaries for both categorical(e.g., user load profile classification) and sequential applications(e.g., renewables generation forecasting). Case studies on classification of power quality disturbances and forecast of building loads demonstrate the vulnerabilities of current ML algorithms in power networks under our adversarial designs. These vulnerabilities call for design of robust and secure ML algorithms for real world applications.

Chi Zhang, Yuanyuan Shi, Yize Chen

Recent advancements in reinforcement learning algorithms have opened doors for researchers to operate and optimize building energy management systems autonomously. However, the lack of an easily configurable building dynamical model and energy management task simulation and evaluation platform has arguably slowed the progress in developing advanced and dedicated reinforcement learning (RL) and control algorithms for building operation tasks. Here we propose "BEAR", a physics-principled Building Environment for Control And Reinforcement Learning. The platform allows researchers to benchmark both model-based and model-free controllers using a broad collection of standard building models in Python without co-simulation using external building simulators. In this paper, we discuss the design of this platform and compare it with other existing building simulation frameworks. We demonstrate the compatibility and performance of BEAR with different controllers, including both model predictive control (MPC) and several state-of-the-art RL methods with two case studies.

Ruohong Liu, Yize Chen

We consider the problem of learning the energy disaggregation signals for residential load data. Such task is referred as non-intrusive load monitoring (NILM), and in order to find individual devices' power consumption profiles based on aggregated meter measurements, a machine learning model is usually trained based on large amount of training data coming from a number of residential homes. Yet collecting such residential load datasets require both huge efforts and customers' approval on sharing metering data, while load data coming from different regions or electricity users may exhibit heterogeneous usage patterns. Both practical concerns make training a single, centralized NILM model challenging. In this paper, we propose a decentralized and task-adaptive learning scheme for NILM tasks, where nested meta learning and federated learning steps are designed for learning task-specific models collectively. Simulation results on benchmark dataset validate proposed algorithm's performance on efficiently inferring appliance-level consumption for a variety of homes and appliances.

Haotian Bai, Yiqi Lin, Yize Chen et al.

The explicit neural radiance field (NeRF) has gained considerable interest for its efficient training and fast inference capabilities, making it a promising direction such as virtual reality and gaming. In particular, PlenOctree (POT)[1], an explicit hierarchical multi-scale octree representation, has emerged as a structural and influential framework. However, POT's fixed structure for direct optimization is sub-optimal as the scene complexity evolves continuously with updates to cached color and density, necessitating refining the sampling distribution to capture signal complexity accordingly. To address this issue, we propose the dynamic PlenOctree DOT, which adaptively refines the sample distribution to adjust to changing scene complexity. Specifically, DOT proposes a concise yet novel hierarchical feature fusion strategy during the iterative rendering process. Firstly, it identifies the regions of interest through training signals to ensure adaptive and efficient refinement. Next, rather than directly filtering out valueless nodes, DOT introduces the sampling and pruning operations for octrees to aggregate features, enabling rapid parameter learning. Compared with POT, our DOT outperforms it by enhancing visual quality, reducing over $55.15$/$68.84\%$ parameters, and providing 1.7/1.9 times FPS for NeRF-synthetic and Tanks $\&$ Temples, respectively. Project homepage:https://vlislab22.github.io/DOT. [1] Yu, Alex, et al. "Plenoctrees for real-time rendering of neural radiance fields." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021.

Mingyuan Yan, Trager Joswig-Jones, Baosen Zhang et al.

Large-scale AI training workloads in modern data centers exhibit rapid and periodic power fluctuations, which may induce significant voltage deviations in power distribution systems. Existing voltage regulation methods, such as droop control, are primarily designed for slowly varying loads and may therefore be ineffective in mitigating these fast fluctuations. In addition, repeated control actions can incur substantial cost. To address this challenge, this paper proposes a decentralized switching-reference voltage control framework that exploits the structured behavior of AI training workloads. We establish conditions for voltage convergence and characterize an effective reference design that aligns with the two dominant operating levels of the AI training workload. The switching rule for voltage references is implemented solely using local voltage measurements, enabling simple local implementation while significantly reducing control effort. Simulation studies demonstrate that the proposed method substantially reduces both voltage deviations and reactive control effort, while remaining compatible with internal data center control strategies without requiring extensive coordination.

Chengyang Gu, Hui Xiong, Yize Chen

Solving real-world optimal control problems are challenging tasks, as the complex, high-dimensional system dynamics are usually unrevealed to the decision maker. It is thus hard to find the optimal control actions numerically. To deal with such modeling and computation challenges, in this paper, we integrate Neural Networks with the Pontryagin's Maximum Principle (PMP), and propose a sample efficient framework NN-PMP-Gradient. The resulting controller can be implemented for systems with unknown and complex dynamics. By taking an iterative approach, the proposed framework not only utilizes the accurate surrogate models parameterized by neural networks, it also efficiently recovers the optimality conditions along with the optimal action sequences via PMP conditions. Numerical simulations on Linear Quadratic Regulator, energy arbitrage of grid-connected lossy battery, control of single pendulum, and two MuJoCo locomotion tasks demonstrate our proposed NN-PMP-Gradient is a general and versatile computation tool for finding optimal solutions. And compared with the widely applied model-free and model-based reinforcement learning (RL) algorithms, our NN-PMP-Gradient achieves higher sample-efficiency and performance in terms of control objectives.

Siwei Wen, Junyan Ye, Peilin Feng et al.

With the rapid advancement of Artificial Intelligence Generated Content (AIGC) technologies, synthetic images have become increasingly prevalent in everyday life, posing new challenges for authenticity assessment and detection. Despite the effectiveness of existing methods in evaluating image authenticity and locating forgeries, these approaches often lack human interpretability and do not fully address the growing complexity of synthetic data. To tackle these challenges, we introduce FakeVLM, a specialized large multimodal model designed for both general synthetic image and DeepFake detection tasks. FakeVLM not only excels in distinguishing real from fake images but also provides clear, natural language explanations for image artifacts, enhancing interpretability. Additionally, we present FakeClue, a comprehensive dataset containing over 100,000 images across seven categories, annotated with fine-grained artifact clues in natural language. FakeVLM demonstrates performance comparable to expert models while eliminating the need for additional classifiers, making it a robust solution for synthetic data detection. Extensive evaluations across multiple datasets confirm the superiority of FakeVLM in both authenticity classification and artifact explanation tasks, setting a new benchmark for synthetic image detection. The code, model weights, and dataset can be found here: https://github.com/opendatalab/FakeVLM.

Ruohong Liu, Yuxin Pan, Yize Chen

Demand flexibility plays a vital role in maintaining grid balance, reducing peak demand, and saving customers' energy bills. Given their highly shiftable load and significant contribution to a building's energy consumption, Heating, Ventilation, and Air Conditioning (HVAC) systems can provide valuable demand flexibility to the power systems by adjusting their energy consumption in response to electricity price and power system needs. To exploit this flexibility in both operation time and power, it is imperative to accurately model and aggregate the load flexibility of a large population of HVAC systems as well as designing effective control algorithms. In this paper, we tackle the curse of dimensionality issue in modeling and control by utilizing the concept of laxity to quantify the emergency level of each HVAC operation request. We further propose a two-level approach to address energy optimization for a large population of HVAC systems. The lower level involves an aggregator to aggregate HVAC load laxity information and use least-laxity-first (LLF) rule to allocate real-time power for individual HVAC systems based on the controller's total power. Due to the complex and uncertain nature of HVAC systems, we leverage a reinforcement learning (RL)-based controller to schedule the total power based on the aggregated laxity information and electricity price. We evaluate the temperature control and energy cost saving performance of a large-scale group of HVAC systems in both single-zone and multi-zone scenarios, under varying climate and electricity market conditions. The experiment results indicate that proposed approach outperforms the centralized methods in the majority of test scenarios, and performs comparably to model-based method in some scenarios.

Yuzhuo Li, Mariam Mughees, Yize Chen et al.

Recent breakthroughs of large language models (LLMs) have exhibited superior capability across major industries and stimulated multi-hundred-billion-dollar investment in AI-centric data centers in the next 3-5 years. This, in turn, bring the increasing concerns on sustainability and AI-related energy usage. However, there is a largely overlooked issue as challenging and critical as AI model and infrastructure efficiency: the disruptive dynamic power consumption behaviour. With fast, transient dynamics, AI infrastructure features ultra-low inertia, sharp power surge and dip, and a significant peak-idle power ratio. The power scale covers from several hundred watts to megawatts, even to gigawatts. These never-seen-before characteristics make AI a very unique load and pose threats to the power grid reliability and resilience. To reveal this hidden problem, this paper examines the scale of AI power consumption, analyzes AI transient behaviour in various scenarios, develops high-level mathematical models to depict AI workload behaviour and discusses the multifaceted challenges and opportunities they potentially bring to existing power grids. Observing the rapidly evolving machine learning (ML) and AI technologies, this work emphasizes the critical need for interdisciplinary approaches to ensure reliable and sustainable AI infrastructure development, and provides a starting point for researchers and practitioners to tackle such challenges.

Yuxin Pan, Yize Chen, Fangzhen Lin

Designing effective policies for the online 3D bin packing problem (3D-BPP) has been a long-standing challenge, primarily due to the unpredictable nature of incoming box sequences and stringent physical constraints. While current deep reinforcement learning (DRL) methods for online 3D-BPP have shown promising results in optimizing average performance over an underlying box sequence distribution, they often fail in real-world settings where some worst-case scenarios can materialize. Standard robust DRL algorithms tend to overly prioritize optimizing the worst-case performance at the expense of performance under normal problem instance distribution. To address these issues, we first introduce a permutation-based attacker to investigate the practical robustness of both DRL-based and heuristic methods proposed for solving online 3D-BPP. Then, we propose an adjustable robust reinforcement learning (AR2L) framework that allows efficient adjustment of robustness weights to achieve the desired balance of the policy's performance in average and worst-case environments. Specifically, we formulate the objective function as a weighted sum of expected and worst-case returns, and derive the lower performance bound by relating to the return under a mixture dynamics. To realize this lower bound, we adopt an iterative procedure that searches for the associated mixture dynamics and improves the corresponding policy. We integrate this procedure into two popular robust adversarial algorithms to develop the exact and approximate AR2L algorithms. Experiments demonstrate that AR2L is versatile in the sense that it improves policy robustness while maintaining an acceptable level of performance for the nominal case.

Xuan He, Honglin Wen, Yufan Zhang et al.

Unit commitment (UC) are essential tools to transmission system operators for finding the most economical and feasible generation schedules and dispatch signals. Constraint screening has been receiving attention as it holds the promise for reducing a number of inactive or redundant constraints in the UC problem, so that the solution process of large scale UC problem can be accelerated by considering the reduced optimization problem. Standard constraint screening approach relies on optimizing over load and generations to find binding line flow constraints, yet the screening is conservative with a large percentage of constraints still reserved for the UC problem. In this paper, we propose a novel machine learning (ML) model to predict the most economical costs given load inputs. Such ML model bridges the cost perspectives of UC decisions to the optimization-based constraint screening model, and can screen out higher proportion of operational constraints. We verify the proposed method's performance on both sample-aware and sample-agnostic setting, and illustrate the proposed scheme can further reduce the computation time on a variety of setup for UC problems.

Haotian Bai, Yize Chen, Lin Wang

Object-centric surface reconstruction from multi-view images is crucial in creating editable digital assets for AR/VR. Due to the lack of geometric constraints, existing methods, e.g., NeuS necessitate annotating the object masks to reconstruct compact surfaces in mesh processing. Mask annotation, however, incurs considerable labor costs due to its cumbersome nature. This paper presents Hi-NeuS, a novel rendering-based framework for neural implicit surface reconstruction, aiming to recover compact and precise surfaces without multi-view object masks. Our key insight is that the overlapping regions in the object-centric views naturally highlight the object of interest as the camera orbits around objects. The object of interest can be specified by estimating the distribution of the rendering weights accumulated from multiple views, which implicitly identifies the surface that a user intends to capture. This inspires us to design a geometric refinement approach, which takes multi-view rendering weights to guide the signed distance functions (SDF) of neural surfaces in a self-supervised manner. Specifically, it retains these weights to resample a pseudo surface based on their distribution. This facilitates the alignment of the SDF to the object of interest. We then regularize the SDF's bias for geometric consistency. Moreover, we propose to use unmasked Chamfer Distance(CD) to measure the extracted mesh without post-processing for more precise evaluation. Our approach has been validated through NeuS and its variant Neuralangelo, demonstrating its adaptability across different NeuS backbones. Extensive benchmark on the DTU dataset shows that our method reduces surface noise by about 20%, and improves the unmasked CD by around 30%, achieving better surface details. The superiority of Hi-NeuS is further validated on BlendedMVS and handheld camera captures for content creation.

Chengyang Gu, Yuxin Pan, Hui Xiong et al.

Model-based reinforcement learning (MBRL) improves sample efficiency by leveraging learned dynamics models for policy optimization. However, the effectiveness of methods such as actor-critic is often limited by compounding model errors, which degrade long-horizon value estimation. Existing approaches, such as Model-Based Value Expansion (MVE), partially mitigate this issue through multi-step rollouts, but remain sensitive to rollout horizon selection and residual model bias. Motivated by the Pontryagin Maximum Principle (PMP), we propose Hamiltonian Actor-Critic (HAC), a model-based approach that eliminates explicit value function learning by directly optimizing a Hamiltonian defined over the learned dynamics and reward for deterministic systems. By avoiding value approximation, HAC reduces sensitivity to model errors while admitting convergence guarantees. Extensive experiments on continuous control benchmarks, in both online and offline RL settings, demonstrate that HAC outperforms model-free and MVE-based baselines in control performance, convergence speed, and robustness to distributional shift, including out-of-distribution (OOD) scenarios. In offline settings with limited data, HAC matches or exceeds state-of-the-art methods, highlighting its strong sample efficiency.

Zhenyang Cai, Junying Chen, Rongsheng Wang et al.

Medical imaging provides essential visual insights for diagnosis, and multimodal large language models (MLLMs) are increasingly utilized for its analysis due to their strong generalization capabilities; however, the underlying factors driving this generalization remain unclear. Current research suggests that multi-task training outperforms single-task as different tasks can benefit each other, but they often overlook the internal relationships within these tasks. To analyze this phenomenon, we attempted to employ compositional generalization (CG), which refers to the models' ability to understand novel combinations by recombining learned elements, as a guiding framework. Since medical images can be precisely defined by Modality, Anatomical area, and Task, naturally providing an environment for exploring CG, we assembled 106 medical datasets to create Med-MAT for comprehensive experiments. The experiments confirmed that MLLMs can use CG to understand unseen medical images and identified CG as one of the main drivers of the generalization observed in multi-task training. Additionally, further studies demonstrated that CG effectively supports datasets with limited data and confirmed that MLLMs can achieve CG across classification and detection tasks, underscoring its broader generalization potential. Med-MAT is available at https://github.com/FreedomIntelligence/Med-MAT.

Shan Jia, Reilin Lyu, Kangran Zhao et al.

DeepFakes, which refer to AI-generated media content, have become an increasing concern due to their use as a means for disinformation. Detecting DeepFakes is currently solved with programmed machine learning algorithms. In this work, we investigate the capabilities of multimodal large language models (LLMs) in DeepFake detection. We conducted qualitative and quantitative experiments to demonstrate multimodal LLMs and show that they can expose AI-generated images through careful experimental design and prompt engineering. This is interesting, considering that LLMs are not inherently tailored for media forensic tasks, and the process does not require programming. We discuss the limitations of multimodal LLMs for these tasks and suggest possible improvements.

Chenghao Huang, Siyang Li, Ruohong Liu et al.

Foundation models, such as Large Language Models (LLMs), can respond to a wide range of format-free queries without any task-specific data collection or model training, creating various research and application opportunities for the modeling and operation of large-scale power systems. In this paper, we outline how such large foundation model such as GPT-4 are developed, and discuss how they can be leveraged in challenging power and energy system tasks. We first investigate the potential of existing foundation models by validating their performance on four representative tasks across power system domains, including the optimal power flow (OPF), electric vehicle (EV) scheduling, knowledge retrieval for power engineering technical reports, and situation awareness. Our results indicate strong capabilities of such foundation models on boosting the efficiency and reliability of power system operational pipelines. We also provide suggestions and projections on future deployment of foundation models in power system applications.

Chengyang Gu, Yuxin Pan, Hui Xiong et al.

Offline reinforcement learning (RL) enables policy learning from pre-collected datasets, avoiding costly and risky online interactions, but it often struggles with long-horizon tasks involving sparse rewards. Existing goal-conditioned and hierarchical offline RL methods decompose such tasks and generate intermediate rewards to mitigate limitations of traditional offline RL, but usually overlook temporal dependencies among subgoals and rely on imprecise reward shaping, leading to suboptimal policies. To address these issues, we propose STO-RL (Offline RL using LLM-Guided Subgoal Temporal Order), an offline RL framework that leverages large language models (LLMs) to generate temporally ordered subgoal sequences and corresponding state-to-subgoal-stage mappings. Using this temporal structure, STO-RL applies potential-based reward shaping to transform sparse terminal rewards into dense, temporally consistent signals, promoting subgoal progress while avoiding suboptimal solutions. The resulting augmented dataset with shaped rewards enables efficient offline training of high-performing policies. Evaluations on four discrete and continuous sparse-reward benchmarks demonstrate that STO-RL consistently outperforms state-of-the-art offline goal-conditioned and hierarchical RL baselines, achieving faster convergence, higher success rates, and shorter trajectories. Ablation studies further confirm STO-RL's robustness to imperfect or noisy LLM-generated subgoal sequences, demonstrating that LLM-guided subgoal temporal structures combined with theoretically grounded reward shaping provide a practical and scalable solution for long-horizon offline RL.

Xun Jiang, Feng Li, Han Zhao et al.

Large language models (LLMs) like GPTs, trained on vast datasets, have demonstrated impressive capabilities in language understanding, reasoning, and planning, achieving human-level performance in various tasks. Most studies focus on enhancing these models by training on ever-larger datasets to build more powerful foundation models. While training stronger models is important, enabling models to evolve during inference is equally crucial, a process we refer to as AI self-evolution. Unlike large-scale training, self-evolution may rely on limited data or interactions. Inspired by the columnar organization of the human cerebral cortex, we hypothesize that AI models could develop cognitive abilities and build internal representations through iterative interactions with their environment. To achieve this, models need long-term memory (LTM) to store and manage processed interaction data. LTM supports self-evolution by representing diverse experiences across environments and agents. In this report, we explore AI self-evolution and its potential to enhance models during inference. We examine LTM's role in lifelong learning, allowing models to evolve based on accumulated interactions. We outline the structure of LTM and the systems needed for effective data retention and representation. We also classify approaches for building personalized models with LTM data and show how these models achieve self-evolution through interaction. Using LTM, our multi-agent framework OMNE achieved first place on the GAIA benchmark, demonstrating LTM's potential for AI self-evolution. Finally, we present a roadmap for future research, emphasizing the importance of LTM for advancing AI technology and its practical applications.

Siyang Li, Hui Xiong, Yize Chen

Due to the vast electric vehicle (EV) penetration to distribution grid, charging load forecasting is essential to promote charging station operation and demand-side management.However, the stochastic charging behaviors and associated exogenous factors render future charging load patterns quite volatile and hard to predict. Accordingly, we devise a novel Diffusion model termed DiffPLF for Probabilistic Load Forecasting of EV charging, which can explicitly approximate the predictive load distribution conditioned on historical data and related covariates. Specifically, we leverage a denoising diffusion model, which can progressively convert the Gaussian prior to real time-series data by learning a reversal of the diffusion process. Besides, we couple such diffusion model with a cross-attention-based conditioning mechanism to execute conditional generation for possible charging demand profiles. We also propose a task-informed fine-tuning technique to better adapt DiffPLF to the probabilistic time-series forecasting task and acquire more accurate and reliable predicted intervals. Finally, we conduct multiple experiments to validate the superiority of DiffPLF to predict complex temporal patterns of erratic charging load and carry out controllable generation based on certain covariate. Results demonstrate that we can attain a notable rise of 39.58% and 49.87% on MAE and CRPS respectively compared to the conventional method.

Yuqi Jiang, Yan Li, Yize Chen

Rapid progress in machine learning and deep learning has enabled a wide range of applications in the electricity load forecasting of power systems, for instance, univariate and multivariate short-term load forecasting. Though the strong capabilities of learning the non-linearity of the load patterns and the high prediction accuracy have been achieved, the interpretability of typical deep learning models for electricity load forecasting is less studied. This paper proposes an interpretable deep learning method, which learns a linear combination of neural networks that each attends to an input time feature. We also proposed a multi-scale time series decomposition method to deal with the complex time patterns. Case studies have been carried out on the Belgium central grid load dataset and the proposed model demonstrated better accuracy compared to the frequently applied baseline model. Specifically, the proposed multi-scale temporal decomposition achieves the best MSE, MAE and RMSE of 0.52, 0.57 and 0.72 respectively. As for interpretability, on one hand, the proposed method displays generalization capability. On the other hand, it can demonstrate not only the feature but also the temporal interpretability compared to other baseline methods. Besides, the global time feature interpretabilities are also obtained. Obtaining global feature interpretabilities allows us to catch the overall patterns, trends, and cyclicality in load data while also revealing the significance of various time-related features in forming the final outputs.

Long Ma, Zhiyuan Yan, Jin Xu et al.

Detecting deepfakes has been an increasingly important topic, especially given the rapid development of AI generation techniques. In this paper, we ask: How can we build a universal detection framework that is effective for most facial deepfakes? One significant challenge is the wide variety of deepfake generators available, resulting in varying forgery artifacts (e.g., lighting inconsistency, color mismatch, etc). But should we ``teach" the detector to learn all these artifacts separately? It is impossible and impractical to elaborate on them all. So the core idea is to pinpoint the more common and general artifacts across different deepfakes. Accordingly, we categorize deepfake artifacts into two distinct yet complementary types: Face Inconsistency Artifacts (FIA) and Up-Sampling Artifacts (USA). FIA arise from the challenge of generating all intricate details, inevitably causing inconsistencies between the complex facial features and relatively uniform surrounding areas. USA, on the other hand, are the inevitable traces left by the generator's decoder during the up-sampling process. This categorization stems from the observation that all existing deepfakes typically exhibit one or both of these artifacts. To achieve this, we propose a new data-level pseudo-fake creation framework that constructs fake samples with only the FIA and USA, without introducing extra less-general artifacts. Specifically, we employ a super-resolution to simulate the USA, while design a Blender module that uses image-level self-blending on diverse facial regions to create the FIA. We surprisingly found that, with this intuitive design, a standard image classifier trained only with our pseudo-fake data can non-trivially generalize well to unseen deepfakes.

Jiayan Nan, Wenquan Ma, Wenlong Wu et al.

Large Language Models (LLMs) demonstrate remarkable capabilities, yet their inability to maintain persistent memory in long contexts limits their effectiveness as autonomous agents in long-term interactions. While existing memory systems have made progress, their reliance on arbitrary granularity for defining the basic memory unit and passive, rule-based mechanisms for knowledge extraction limits their capacity for genuine learning and evolution. To address these foundational limitations, we present Nemori, a novel self-organizing memory architecture inspired by human cognitive principles. Nemori's core innovation is twofold: First, its Two-Step Alignment Principle, inspired by Event Segmentation Theory, provides a principled, top-down method for autonomously organizing the raw conversational stream into semantically coherent episodes, solving the critical issue of memory granularity. Second, its Predict-Calibrate Principle, inspired by the Free-energy Principle, enables the agent to proactively learn from prediction gaps, moving beyond pre-defined heuristics to achieve adaptive knowledge evolution. This offers a viable path toward handling the long-term, dynamic workflows of autonomous agents. Extensive experiments on the LoCoMo and LongMemEval benchmarks demonstrate that Nemori significantly outperforms prior state-of-the-art systems, with its advantage being particularly pronounced in longer contexts.

Siyang Li, Yize Chen, Zijie Zhu et al.

Time Series Foundation Models (TSFMs) advance generalization and data efficiency in time series forecasting by unified large-scale pretraining. But TSFMs remain lacking when adapting to specific downstream forecasting tasks for two reasons. First, the non-stationary and uncertain nature of time series data lead to inevitable temporal distribution shifts between historical training and future testing data, while current Supervised FineTuning (SFT)-based methods are prone to overfitting and may degrade generalization. Second, training data availability varies across forecasting tasks, requiring TSFMs to generalize well under diverse data regimes. To address these challenges, we introduce the Time series Reinforcement Finetuning (TimeRFT) paradigm for TSFM downstream adaptation, which consists of two task-specific training recipes: i) A forecasting quality-based temporal reward mechanism that conducts a multi-faceted evaluation of the contribution of each prediction step to overall forecasting accuracy. ii) A forecasting difficulty-based data selection strategy to identify time series samples with generalizable predictive patterns and informative training signals. Extensive experiments demonstrate TimeRFT can consistently outperform SFT-based adaptation methods across various real-world forecasting tasks and training data regimes, enhancing prediction accuracy and generalization against unforeseen distribution shifts.

Yuxin Pan, Ruohong Liu, Yize Chen et al.

Neural solvers based on the divide-and-conquer approach for Vehicle Routing Problems (VRPs) in general, and capacitated VRP (CVRP) in particular, integrates the global partition of an instance with local constructions for each subproblem to enhance generalization. However, during the global partition phase, misclusterings within subgraphs have a tendency to progressively compound throughout the multi-step decoding process of the learning-based partition policy. This suboptimal behavior in the global partition phase, in turn, may lead to a dramatic deterioration in the performance of the overall decomposition-based system, despite using optimal local constructions. To address these challenges, we propose a versatile Hierarchical Learning-based Graph Partition (HLGP) framework, which is tailored to benefit the partition of CVRP instances by synergistically integrating global and local partition policies. Specifically, the global partition policy is tasked with creating the coarse multi-way partition to generate the sequence of simpler two-way partition subtasks. These subtasks mark the initiation of the subsequent K local partition levels. At each local partition level, subtasks exclusive for this level are assigned to the local partition policy which benefits from the insensitive local topological features to incrementally alleviate the compounded errors. This framework is versatile in the sense that it optimizes the involved partition policies towards a unified objective harmoniously compatible with both reinforcement learning (RL) and supervised learning (SL). (*Due to the notification of arXiv "The Abstract field cannot be longer than 1,920 characters", the appeared Abstract is shortened. For the full Abstract, please download the Article.)

Yize Chen, Xiaogui Zheng

The integration of AI data centers into power grid represents one of the most emerging and complex challenges for the energy systems. As computational demand scales at an unprecedented rate, the traditional grid planning study's paradigm of treating data centers as rigid, inflexible loads is becoming economically, mathematically and operationally untenable. This work tries to understand and address the large load interconnection bottleneck by modeling and evaluating AI load flexibility. By examining data center's temporal and spatial shifting capabilities within a grid capacity expansion framework, we build a quantitative grid planning model, and evaluate their impacts on additional generation, operational costs, and network congestion. Numerical study reveals interesting observations, as AI data center flexibility are not felt consistently, and increasing flexibility does not necessarily translate to less generation capacity required. Depending on data center's locations, flexibility range, and grid load conditions, flexible AI load can help reduce grid investment and operational costs by 3-21%. Our work also indicate that longer deferral time of AI compute has diminishing returns for offloading grid electricity dispatch pressure.

Chengyang Gu, Yuxin Pan, Ruohong Liu et al.

In the context of charging electric vehicles (EVs), the price-based demand response (PBDR) is becoming increasingly significant for charging load management. Such response usually encourages cost-sensitive customers to adjust their energy demand in response to changes in price for financial incentives. Thus, to model and optimize EV charging, it is important for charging station operator to model the PBDR patterns of EV customers by precisely predicting charging demands given price signals. Then the operator refers to these demands to optimize charging station power allocation policy. The standard pipeline involves offline fitting of a PBDR function based on historical EV charging records, followed by applying estimated EV demands in downstream charging station operation optimization. In this work, we propose a new decision-focused end-to-end framework for PBDR modeling that combines prediction errors and downstream optimization cost errors in the model learning stage. We evaluate the effectiveness of our method on a simulation of charging station operation with synthetic PBDR patterns of EV customers, and experimental results demonstrate that this framework can provide a more reliable prediction model for the ultimate optimization process, leading to more effective optimization solutions in terms of cost savings and charging station operation objectives with only a few training samples.

Kangran Zhao, Yupeng Chen, Xiaoyu Zhang et al.

The misuse of advanced generative AI models has resulted in the widespread proliferation of falsified data, particularly forged human-centric audiovisual content, which poses substantial societal risks (e.g., financial fraud and social instability). In response to this growing threat, several works have preliminarily explored countermeasures. However, the lack of sufficient and diverse training data, along with the absence of a standardized benchmark, hinder deeper exploration. To address this challenge, we first build Mega-MMDF, a large-scale, diverse, and high-quality dataset for multimodal deepfake detection. Specifically, we employ 21 forgery pipelines through the combination of 10 audio forgery methods, 12 visual forgery methods, and 6 audio-driven face reenactment methods. Mega-MMDF currently contains 0.1 million real samples and 1.1 million forged samples, making it one of the largest and most diverse multimodal deepfake datasets, with plans for continuous expansion. Building on it, we present DeepfakeBench-MM, the first unified benchmark for multimodal deepfake detection. It establishes standardized protocols across the entire detection pipeline and serves as a versatile platform for evaluating existing methods as well as exploring novel approaches. DeepfakeBench-MM currently supports 5 datasets and 11 multimodal deepfake detectors. Furthermore, our comprehensive evaluations and in-depth analyses uncover several key findings from multiple perspectives (e.g., augmentation, stacked forgery). We believe that DeepfakeBench-MM, together with our large-scale Mega-MMDF, will serve as foundational infrastructures for advancing multimodal deepfake detection.

Yuxin Pan, Zhiguang Cao, Chengyang Gu et al.

Existing neural methods for multi-task vehicle routing problems (VRPs) typically learn unified solvers to handle multiple constraints simultaneously. However, they often underutilize the compositional structure of VRP variants, each derivable from a common set of basis VRP variants. This critical oversight causes unified solvers to miss out the potential benefits of basis solvers, each specialized for a basis VRP variant. To overcome this limitation, we propose a framework that enables unified solvers to perceive the shared-component nature across VRP variants by proactively reusing basis solvers, while mitigating the exponential growth of trained neural solvers. Specifically, we introduce a State-Decomposable MDP (SDMDP) that reformulates VRPs by expressing the state space as the Cartesian product of basis state spaces associated with basis VRP variants. More crucially, this formulation inherently yields the optimal basis policy for each basis VRP variant. Furthermore, a Latent Space-based SDMDP extension is developed by incorporating both the optimal basis policies and a learnable mixture function to enable the policy reuse in the latent space. Under mild assumptions, this extension provably recovers the optimal unified policy of SDMDP through the mixture function that computes the state embedding as a mapping from the basis state embeddings generated by optimal basis policies. For practical implementation, we introduce the Mixture-of-Specialized-Experts Solver (MoSES), which realizes basis policies through specialized Low-Rank Adaptation (LoRA) experts, and implements the mixture function via an adaptive gating mechanism. Extensive experiments conducted across VRP variants showcase the superiority of MoSES over prior methods.

Siyang Li, Yize Chen, Yan Guo et al.

Advanced deep learning-based approaches have been actively applied to forecast the spatiotemporal physical dynamics governed by partial differential equations (PDEs), which acts as a critical procedure in tackling many science and engineering problems. As real-world physical environments like PDE system parameters are always capricious, how to generalize across unseen out-of-distribution (OOD) forecasting scenarios using limited training data is of great importance. To bridge this barrier, existing methods focus on discovering domain-generalizable representations across various PDE dynamics trajectories. However, their zero-shot OOD generalization capability remains deficient, since extra test-time samples for domain-specific adaptation are still required. This is because the fundamental physical invariance in PDE dynamical systems are yet to be investigated or integrated. To this end, we first explicitly define a two-fold PDE invariance principle, which points out that ingredient operators and their composition relationships remain invariant across different domains and PDE system evolution. Next, to capture this two-fold PDE invariance, we propose a physics-guided invariant learning method termed iMOOE, featuring an Invariance-aligned Mixture Of Operator Expert architecture and a frequency-enriched invariant learning objective. Extensive experiments across simulated benchmarks and real-world applications validate iMOOE's superior in-distribution performance and zero-shot generalization capabilities on diverse OOD forecasting scenarios.

Ruohong Liu, Jack Umenberger, Yize Chen

Recent years have seen significant advancements in designing reinforcement learning (RL)-based agents for building energy management. While individual success is observed in simulated or controlled environments, the scalability of RL approaches in terms of efficiency and generalization across building dynamics and operational scenarios remains an open question. In this work, we formally characterize the generalization space for the cross-environment, multi-objective building energy management task, and formulate the multi-objective contextual RL problem. Such a formulation helps understand the challenges of transferring learned policies across varied operational contexts such as climate and heat convection dynamics under multiple control objectives such as comfort level and energy consumption. We provide a principled way to parameterize such contextual information in realistic building RL environments, and construct a novel benchmark to facilitate the evaluation of generalizable RL algorithms in practical building control tasks. Our results show that existing multi-objective RL methods are capable of achieving reasonable trade-offs between conflicting objectives. However, their performance degrades under certain environment variations, underscoring the importance of incorporating dynamics-dependent contextual information into the policy learning process.

Alexander Pan, Yongkyun Lee, Huan Zhang et al.

Due to the proliferation of renewable energy and its intrinsic intermittency and stochasticity, current power systems face severe operational challenges. Data-driven decision-making algorithms from reinforcement learning (RL) offer a solution towards efficiently operating a clean energy system. Although RL algorithms achieve promising performance compared to model-based control models, there has been limited investigation of RL robustness in safety-critical physical systems. In this work, we first show that several competition-winning, state-of-the-art RL agents proposed for power system control are vulnerable to adversarial attacks. Specifically, we use an adversary Markov Decision Process to learn an attack policy, and demonstrate the potency of our attack by successfully attacking multiple winning agents from the Learning To Run a Power Network (L2RPN) challenge, under both white-box and black-box attack settings. We then propose to use adversarial training to increase the robustness of RL agent against attacks and avoid infeasible operational decisions. To the best of our knowledge, our work is the first to highlight the fragility of grid control RL algorithms, and contribute an effective defense scheme towards improving their robustness and security.

Congmei Jiang, Yize Chen, Yongfang Mao et al.

The operation and planning of large-scale power systems are becoming more challenging with the increasing penetration of stochastic renewable generation. In order to minimize the decision risks in power systems with large amount of renewable resources, there is a growing need to model the short-term generation uncertainty. By producing a group of possible future realizations for certain set of renewable generation plants, scenario approach has become one popular way for renewables uncertainty modeling. However, due to the complex spatial and temporal correlations underlying in renewable generations, traditional model-based approaches for forecasting future scenarios often require extensive knowledge, while fitted models are often hard to scale. To address such modeling burdens, we propose a learning-based, data-driven scenario forecasts method based on generative adversarial networks (GANs), which is a class of deep-learning generative algorithms used for modeling unknown distributions. We firstly utilize an improved GANs with convergence guarantees to learn the intrinsic patterns and model the unknown distributions of (multiple-site) renewable generation time-series. Then by solving an optimization problem, we are able to generate forecasted scenarios without any scenario number and forecasting horizon restrictions. Our method is totally model-free, and could forecast scenarios under different level of forecast uncertainties. Extensive numerical simulations using real-world data from NREL wind and solar integration datasets validate the performance of proposed method in forecasting both wind and solar power scenarios.

Yize Chen, Yushi Tan, Baosen Zhang

Load forecasting plays a critical role in the operation and planning of power systems. By using input features such as historical loads and weather forecasts, system operators and utilities build forecast models to guide decision making in commitment and dispatch. As the forecasting techniques becomes more sophisticated, however, they also become more vulnerable to cybersecurity threats. In this paper, we study the vulnerability of a class of load forecasting algorithms and analyze the potential impact on the power system operations, such as load shedding and increased dispatch costs. Specifically, we propose data injection attack algorithms that require minimal assumptions on the ability of the adversary. The attacker does not need to have knowledge about the load forecasting model or the underlying power system. Surprisingly, our results indicate that standard load forecasting algorithms are quite vulnerable to the designed black-box attacks. By only injecting malicious data in temperature from online weather forecast APIs, an attacker could manipulate load forecasts in arbitrary directions and cause significant and targeted damages to system operations.

Yize Chen, Hao Wang

The prosperity of smart mobile devices has made mobile crowdsensing (MCS) a promising paradigm for completing complex sensing and computation tasks. In the past, great efforts have been made on the design of incentive mechanisms and task allocation strategies from MCS platform's perspective to motivate mobile users' participation. However, in practice, MCS participants face many uncertainties coming from their sensing environment as well as other participants' strategies, and how do they interact with each other and make sensing decisions is not well understood. In this paper, we take MCS participants' perspective to derive an online sensing policy to maximize their payoffs via MCS participation. Specifically, we model the interactions of mobile users and sensing environments as a multi-agent Markov decision process. Each participant cannot observe others' decisions, but needs to decide her effort level in sensing tasks only based on local information, e.g., its own record of sensed signals' quality. To cope with the stochastic sensing environment, we develop an intelligent crowdsensing algorithm IntelligentCrowd by leveraging the power of multi-agent reinforcement learning (MARL). Our algorithm leads to the optimal sensing policy for each user to maximize the expected payoff against stochastic sensing environments, and can be implemented at individual participant's level in a distributed fashion. Numerical simulations demonstrate that IntelligentCrowd significantly improves users' payoffs in sequential MCS tasks under various sensing dynamics.

Yize Chen, Pan Li, Baosen Zhang

We present a method to generate renewable scenarios using Bayesian probabilities by implementing the Bayesian generative adversarial network~(Bayesian GAN), which is a variant of generative adversarial networks based on two interconnected deep neural networks. By using a Bayesian formulation, generators can be constructed and trained to produce scenarios that capture different salient modes in the data, allowing for better diversity and more accurate representation of the underlying physical process. Compared to conventional statistical models that are often hard to scale or sample from, this method is model-free and can generate samples extremely efficiently. For validation, we use wind and solar times-series data from NREL integration data sets to train the Bayesian GAN. We demonstrate that proposed method is able to generate clusters of wind scenarios with different variance and mean value, and is able to distinguish and generate wind and solar scenarios simultaneously even if the historical data are intentionally mixed.

Yize Chen, Yishen Wang, Daniel Kirschen et al.

Scenario generation is an important step in the operation and planning of power systems with high renewable penetrations. In this work, we proposed a data-driven approach for scenario generation using generative adversarial networks, which is based on two interconnected deep neural networks. Compared with existing methods based on probabilistic models that are often hard to scale or sample from, our method is data-driven, and captures renewable energy production patterns in both temporal and spatial dimensions for a large number of correlated resources. For validation, we use wind and solar times-series data from NREL integration data sets. We demonstrate that the proposed method is able to generate realistic wind and photovoltaic power profiles with full diversity of behaviors. We also illustrate how to generate scenarios based on different conditions of interest by using labeled data during training. For example, scenarios can be conditioned on weather events~(e.g. high wind day) or time of the year~(e,g. solar generation for a day in July). Because of the feedforward nature of the neural networks, scenarios can be generated extremely efficiently without sophisticated sampling techniques.

Hossein Hosseini, Yize Chen, Sreeram Kannan et al.

Advances in Machine Learning (ML) have led to its adoption as an integral component in many applications, including banking, medical diagnosis, and driverless cars. To further broaden the use of ML models, cloud-based services offered by Microsoft, Amazon, Google, and others have developed ML-as-a-service tools as black-box systems. However, ML classifiers are vulnerable to adversarial examples: inputs that are maliciously modified can cause the classifier to provide adversary-desired outputs. Moreover, it is known that adversarial examples generated on one classifier are likely to cause another classifier to make the same mistake, even if the classifiers have different architectures or are trained on disjoint datasets. This property, which is known as transferability, opens up the possibility of attacking black-box systems by generating adversarial examples on a substitute classifier and transferring the examples to the target classifier. Therefore, the key to protect black-box learning systems against the adversarial examples is to block their transferability. To this end, we propose a training method that, as the input is more perturbed, the classifier smoothly outputs lower confidence on the original label and instead predicts that the input is "invalid". In essence, we augment the output class set with a NULL label and train the classifier to reject the adversarial examples by classifying them as NULL. In experiments, we apply a wide range of attacks based on adversarial examples on the black-box systems. We show that a classifier trained with the proposed method effectively resists against the adversarial examples, while maintaining the accuracy on clean data.