arXiv subjects

Mingxuan Li

Publications and source records attributed to Mingxuan Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

AmbiBench: Benchmarking Mobile GUI Agents Beyond One-Shot Instructions in the Wild

Benchmarks are paramount for gauging progress in the domain of Mobile GUI Agents. In practical scenarios, users frequently fail to articulate precise directives containing full task details at the onset, and their expressions are typically ambiguous. Consequently, agents are required to converge on the user's true intent via active clarification and interaction during execution. However, existing benchmarks predominantly operate under the idealized assumption that user-issued instructions are complete and unequivocal. This paradigm focuses exclusively on assessing single-turn execution while overlooking the alignment capability of the agent. To address this limitation, we introduce AmbiBench, the first benchmark incorporating a taxonomy of instruction clarity to shift evaluation from unidirectional instruction following to bidirectional intent alignment. Grounded in Cognitive Gap theory, we propose a taxonomy of four clarity levels: Detailed, Standard, Incomplete, and Ambiguous. We construct a rigorous dataset of 240 ecologically valid tasks across 25 applications, subject to strict review protocols. Furthermore, targeting evaluation in dynamic environments, we develop MUSE (Mobile User Satisfaction Evaluator), an automated framework utilizing an MLLM-as-a-judge multi-agent architecture. MUSE performs fine-grained auditing across three dimensions: Outcome Effectiveness, Execution Quality, and Interaction Quality. Empirical results on AmbiBench reveal the performance boundaries of SoTA agents across different clarity levels, quantify the gains derived from active interaction, and validate the strong correlation between MUSE and human judgment. This work redefines evaluation standards, laying the foundation for next-generation agents capable of truly understanding user intent.

cs.SE↗

arXiv2026

DR.Experts: Differential Refinement of Distortion-Aware Experts for Blind Image Quality Assessment

Blind Image Quality Assessment, aiming to replicate human perception of visual quality without reference, plays a key role in vision tasks, yet existing models often fail to effectively capture subtle distortion cues, leading to a misalignment with human subjective judgments. We identify that the root cause of this limitation lies in the lack of reliable distortion priors, as methods typically learn shallow relationships between unified image features and quality scores, resulting in their insensitive nature to distortions and thus limiting their performance. To address this, we introduce DR.Experts, a novel prior-driven BIQA framework designed to explicitly incorporate distortion priors, enabling a reliable quality assessment. DR.Experts begins by leveraging a degradation-aware vision-language model to obtain distortion-specific priors, which are further refined and enhanced by the proposed Distortion-Saliency Differential Module through distinguishing them from semantic attentions, thereby ensuring the genuine representations of distortions. The refined priors, along with semantics and bridging representation, are then fused by a proposed mixture-of-experts style module named the Dynamic Distortion Weighting Module. This mechanism weights each distortion-specific feature as per its perceptual impact, ensuring that the final quality prediction aligns with human perception. Extensive experiments conducted on five challenging BIQA benchmarks demonstrate the superiority of DR.Experts over current methods and showcase its excellence in terms of generalization and data efficiency.

cs.CV↗

arXiv2026

Confounding Robust Continuous Control via Automatic Reward Shaping

Reward shaping has been applied widely to accelerate Reinforcement Learning (RL) agents' training. However, a principled way of designing effective reward shaping functions, especially for complex continuous control problems, remains largely under-explained. In this work, we propose to automatically learn a reward shaping function for continuous control problems from offline datasets, potentially contaminated by unobserved confounding variables. Specifically, our method builds upon the recently proposed causal Bellman equation to learn a tight upper bound on the optimal state values, which is then used as the potentials in the Potential-Based Reward Shaping (PBRS) framework. Our proposed reward shaping algorithm is tested with Soft-Actor-Critic (SAC) on multiple commonly used continuous control benchmarks and exhibits strong performance guarantees under unobserved confounders. More broadly, our work marks a solid first step towards confounding robust continuous control from a causal perspective. Code for training our reward shaping functions can be found at https://github.com/mateojuliani/confounding_robust_cont_control.

cs.LG↗

arXiv2026

Causal Flow Q-Learning for Robust Offline Reinforcement Learning

Expressive policies based on flow-matching have been successfully applied in reinforcement learning (RL) more recently due to their ability to model complex action distributions from offline data. These algorithms build on standard policy gradients, which assume that there is no unmeasured confounding in the data. However, this condition does not necessarily hold for pixel-based demonstrations when a mismatch exists between the demonstrator's and the learner's sensory capabilities, leading to implicit confounding biases in offline data. We address the challenge by investigating the problem of confounded observations in offline RL from a causal perspective. We develop a novel causal offline RL objective that optimizes policies' worst-case performance that may arise due to confounding biases. Based on this new objective, we introduce a practical implementation that learns expressive flow-matching policies from confounded demonstrations, employing a deep discriminator to assess the discrepancy between the target policy and the nominal behavioral policy. Experiments across 25 pixel-based tasks demonstrate that our proposed confounding-robust augmentation procedure achieves a success rate 120\% that of confounding-unaware, state-of-the-art offline RL methods.

cs.LG↗

arXiv2026

Post-Earthquake Restoration of Electricity-Gas Distribution Systems with Damage Information Collection and Repair Vehicle Routing

Extreme events such as earthquakes pose significant threats to integrated electricity-gas distribution systems (IEGDS) by causing widespread damage. Existing restoration approaches typically assume full awareness of damage, which may not be true if monitoring and communication infrastructures are impaired. In such circumstances, field inspection is necessary. This paper presents a novel adaptive restoration framework for IEGDS, considering dynamic damage assessment and repair. The restoration problem is formulated as a partially observable Markov decision process (POMDP), capturing the gradually revealed contingency and the evolving impact of field crew actions. To address the computational challenges of POMDPs in real-time applications, an advanced belief tree search (BTS) algorithm is introduced. This algorithm enables crew members to continuously update their actions based on evolving belief states, leveraging comprehensive simulations to evaluate potential future trajectories and identify optimal inspection and repair strategies. Based on the BTS algorithm, a unified real-time decision-making framework is developed for IEGDS restoration. Case studies on two distinct IEGDS systems demonstrate the effectiveness and scalability of the proposed method. The results indicate that the proposed approach achieves an outage cost comparable to the ideal solution, and reduces the total outage cost by more than 15% compared to strategies based on stochastic programming and heuristic methods.

eess.SY↗

arXiv2025

Distribution Network Restoration with Mobile Resources Dispatch: A Simulation-Based Online Dynamic Programming Approach

Dispatching mobile resources such as repair crews and mobile emergency generators is essential for the rapid restoration of distribution systems after extreme events. However, the restoration process is affected by various uncertain factors including repair time, road condition, and newly observed failures, necessitating online decision-making in response to real-time information. This paper proposes a simulation-based online dynamic programming approach to provide real-time restoration actions considering the dispatch of mobile resources. Using an index-based priority rule as the base policy, the remaining cumulative loss at the current state and a given action is evaluated from online simulation. As the base policy is explicit, the simulation is efficient. Then, the action leading to the minimum cumulative loss is chosen. It is proven that such a strategy improves the performance of base policy. The proposed policy adapts to real-time information updates and does not rely on offline training, so incurs no data and convergence-related issues, which is important in restoration tasks where the historical data of extreme events is rare. The rolling optimization approach may not meet the requirement of online use, because routing mobile resources gives rise to large-scale discrete optimization problems. Case studies on 123-bus and 8500-bus systems demonstrate that the proposed method achieves higher efficiency and better performance compared with rolling horizon optimization.

eess.SY↗

arXiv2025

Robust, Observable, and Evolvable Agentic Systems Engineering: A Principled Framework Validated via the Fairy GUI Agent

The Agentic Paradigm faces a significant Software Engineering Absence, yielding Agentic systems commonly lacking robustness, observability, and evolvability. To address these deficiencies, we propose a principled engineering framework comprising Runtime Goal Refinement (RGR), Observable Cognitive Architecture (OCA), and Evolutionary Memory Architecture (EMA). In this framework, RGR ensures robustness and intent alignment via knowledge-constrained refinement and human-in-the-loop clarification; OCA builds an observable and maintainable white-box architecture using component decoupling, logic layering, and state-control separation; and EMA employs an execution-evolution dual-loop for evolvability. We implemented and empirically validated Fairy, a mobile GUI agent based on this framework. On RealMobile-Eval, our novel benchmark for ambiguous and complex tasks, Fairy outperformed the best SoTA baseline in user requirement completion by 33.7%. Subsequent controlled experiments, human-subject studies, and ablation studies further confirmed that the RGR enhances refinement accuracy and prevents intent deviation; the OCA improves maintainability; and the EMA is crucial for long-term performance. This research provides empirically validated specifications and a practical blueprint for building reliable, observable, and evolvable Agentic AI systems.

cs.AI↗

arXiv2025

IFG: Internet-Scale Guidance for Functional Grasping Generation

Large Vision Models trained on internet-scale data have demonstrated strong capabilities in segmenting and semantically understanding object parts, even in cluttered, crowded scenes. However, while these models can direct a robot toward the general region of an object, they lack the geometric understanding required to precisely control dexterous robotic hands for 3D grasping. To overcome this, our key insight is to leverage simulation with a force-closure grasping generation pipeline that understands local geometries of the hand and object in the scene. Because this pipeline is slow and requires ground-truth observations, the resulting data is distilled into a diffusion model that operates in real-time on camera point clouds. By combining the global semantic understanding of internet-scale models with the geometric precision of a simulation-based locally-aware force-closure, \our achieves high-performance semantic grasping without any manually collected training data. For visualizations of this please visit our website at https://ifgrasping.github.io/

cs.RO↗

arXiv2025

Solvaformer: an SE(3)-equivariant graph transformer for small molecule solubility prediction

Accurate prediction of small molecule solubility using material-sparing approaches is critical for accelerating synthesis and process optimization, yet experimental measurement is costly and many learning approaches either depend on quantumderived descriptors or offer limited interpretability. We introduce Solvaformer, a geometry-aware graph transformer that models solutions as multiple molecules with independent SE(3) symmetries. The architecture combines intramolecular SE(3)-equivariant attention with intermolecular scalar attention, enabling cross-molecular communication without imposing spurious relative geometry. We train Solvaformer in a multi-task setting to predict both solubility (log S) and solvation free energy, using an alternating-batch regimen that trains on quantum-mechanical data (CombiSolv-QM) and on experimental measurements (BigSolDB 2.0). Solvaformer attains the strongest overall performance among the learned models and approaches a DFT-assisted gradient-boosting baseline, while outperforming an EquiformerV2 ablation and sequence-based alternatives. In addition, token-level attention produces chemically coherent attributions: case studies recover known intra- vs. inter-molecular hydrogen-bonding patterns that govern solubility differences in positional isomers. Taken together, Solvaformer provides an accurate, scalable, and interpretable approach to solution-phase property prediction by uniting geometric inductive bias with a mixed dataset training strategy on complementary computational and experimental data.

physics.chem-ph↗

arXiv2025

Confounding Robust Deep Reinforcement Learning: A Causal Approach

A key task in Artificial Intelligence is learning effective policies for controlling agents in unknown environments to optimize performance measures. Off-policy learning methods, like Q-learning, allow learners to make optimal decisions based on past experiences. This paper studies off-policy learning from biased data in complex and high-dimensional domains where \emph{unobserved confounding} cannot be ruled out a priori. Building on the well-celebrated Deep Q-Network (DQN), we propose a novel deep reinforcement learning algorithm robust to confounding biases in observed data. Specifically, our algorithm attempts to find a safe policy for the worst-case environment compatible with the observations. We apply our method to twelve confounded Atari games, and find that it consistently dominates the standard DQN in all games where the observed input to the behavioral and target policies mismatch and unobserved confounders exist.

cs.AI↗

arXiv2025

Automatic Reward Shaping from Confounded Offline Data

cs.AI↗

arXiv2025

Spatiotemporal Prediction of Electric Vehicle Charging Load Based on Large Language Models

The rapid growth of EVs and the subsequent increase in charging demand pose significant challenges for load grid scheduling and the operation of EV charging stations. Effectively harnessing the spatiotemporal correlations among EV charging stations to improve forecasting accuracy is complex. To tackle these challenges, we propose EV-LLM for EV charging loads based on LLMs in this paper. EV-LLM integrates the strengths of Graph Convolutional Networks (GCNs) in spatiotemporal feature extraction with the generalization capabilities of fine-tuned generative LLMs. Also, EV-LLM enables effective data mining and feature extraction across multimodal and multidimensional datasets, incorporating historical charging data, weather information, and relevant textual descriptions to enhance forecasting accuracy for multiple charging stations. We validate the effectiveness of EV-LLM by using charging data from 10 stations in California, demonstrating its superiority over the other traditional deep learning methods and potential to optimize load grid scheduling and support vehicle-to-grid interactions.

eess.SP↗

arXiv2025

Enforced Interface Constraints for Domain Decomposition Method of Discrete Physics-Informed Neural Networks

This study presents a discrete physics-informed neural network (dPINN) framework, enhanced with enforced interface constraints (EIC), for modeling physical systems using the domain decomposition method (DDM). Built upon finite element-style mesh discretization, the dPINN accurately evaluates system energy through Gaussian quadrature-based element-wise integration. To ensure physical field continuity across subdomain interfaces, the EIC mechanism enforces interfacial displacement constraints without requiring auxiliary sampling or loss penalties.This formulation supports independent meshing in each subdomain, simplifying preprocessing and improving computational flexibility. Additionally, by eliminating the influence of weak spatial constraints (WSC) commonly observed in traditional PINNs, the EIC-dPINN delivers more stable and physically consistent predictions.Extensive two- and three-dimensional numerical experiments validate the proposed framework's accuracy and demonstrate the computational efficiency gains achieved through parallel training. The results highlight the framework's scalability, robustness, and potential for solving large-scale, geometrically complex problems.

cs.CE↗

arXiv2025

Photoionization time delays probe electron correlations

The photoelectric effect, explained by Einstein in 1905, is often regarded as a one-electron phenomenon. However, in multi-electron systems, the interaction of the escaping electron with other electrons, referred to as electron correlation, plays an important role. For example, electron correlations in photoionization of the outer $s$-subshells of rare gas atoms lead to a substantial minimum in the ionization probability, which was theoretically predicted in 1972 and experimentally confirmed using synchrotron radiation. However, recent attosecond photoionization time delay measurements in argon strongly disagree with theory, thus raising questions on the nature of electron correlations leading to this minimum. In this work, combining high-spectral resolution attosecond interferometry experiments and novel theoretical calculations allows us to identify the most essential electron correlations affecting the photoemission. The measurement of time delays gives unprecedented insight into the photoionization process, unraveling details of the atomic potential experienced by the escaping electron and capturing its dynamics.

physics.atom-ph↗

arXiv2025

HypoEval: Hypothesis-Guided Evaluation for Natural Language Generation

Large language models (LLMs) have demonstrated great potential for automating the evaluation of natural language generation. Previous frameworks of LLM-as-a-judge fall short in two ways: they either use zero-shot setting without consulting any human input, which leads to low alignment, or fine-tune LLMs on labeled data, which requires a non-trivial number of samples. Moreover, previous methods often provide little reasoning behind automated evaluations. In this paper, we propose HypoEval, Hypothesis-guided Evaluation framework, which first uses a small corpus of human evaluations to generate more detailed rubrics for human judgments and then incorporates a checklist-like approach to combine LLM's assigned scores on each decomposed dimension to acquire overall scores. With only 30 human evaluations, HypoEval achieves state-of-the-art performance in alignment with both human rankings (Spearman correlation) and human scores (Pearson correlation), on average outperforming G-Eval by 11.86% and fine-tuned Llama-3.1-8B-Instruct with at least 3 times more human evaluations by 11.95%. Furthermore, we conduct systematic studies to assess the robustness of HypoEval, highlighting its effectiveness as a reliable and interpretable automated evaluation framework.

cs.CL↗

arXiv2025

Causally Aligned Curriculum Learning

A pervasive challenge in Reinforcement Learning (RL) is the "curse of dimensionality" which is the exponential growth in the state-action space when optimizing a high-dimensional target task. The framework of curriculum learning trains the agent in a curriculum composed of a sequence of related and more manageable source tasks. The expectation is that when some optimal decision rules are shared across source tasks and the target task, the agent could more quickly pick up the necessary skills to behave optimally in the environment, thus accelerating the learning process. However, this critical assumption of invariant optimal decision rules does not necessarily hold in many practical applications, specifically when the underlying environment contains unobserved confounders. This paper studies the problem of curriculum RL through causal lenses. We derive a sufficient graphical condition characterizing causally aligned source tasks, i.e., the invariance of optimal decision rules holds. We further develop an efficient algorithm to generate a causally aligned curriculum, provided with qualitative causal knowledge of the target task. Finally, we validate our proposed methodology through experiments in discrete and continuous confounded tasks with pixel observations.

cs.LG↗

arXiv2025

Modeling, Simulation, and Application of Spatio-Temporal Characteristics Detection in Incipient Slip

Incipient slip detection provides critical feedback for robotic grasping and manipulation tasks. However, maintaining its adaptability under diverse object properties and complex working conditions remains challenging. This article highlights the importance of completely representing spatio-temporal features of slip, and proposes a novel approach for incipient slip modeling and detection. Based on the analysis of localized displacement phenomenon, we establish the relationship between the characteristic strain rate extreme events and the local slip state. This approach enables the detection of both the spatial distribution and temporal dynamics of stick-slip regions. Also, the proposed method can be applied to strain distribution sensing devices, such as vision-based tactile sensors. Simulations and prototype experiments validated the effectiveness of this approach under varying contact conditions, including different contact geometries, friction coefficients, and combined loads. Experiments demonstrated that this method not only accurately and reliably delineates incipient slip, but also facilitates friction parameter estimation and adaptive grasping control.

cs.RO↗

arXiv2025

Literature Meets Data: A Synergistic Approach to Hypothesis Generation

AI holds promise for transforming scientific processes, including hypothesis generation. Prior work on hypothesis generation can be broadly categorized into theory-driven and data-driven approaches. While both have proven effective in generating novel and plausible hypotheses, it remains an open question whether they can complement each other. To address this, we develop the first method that combines literature-based insights with data to perform LLM-powered hypothesis generation. We apply our method on five different datasets and demonstrate that integrating literature and data outperforms other baselines (8.97\% over few-shot, 15.75\% over literature-based alone, and 3.37\% over data-driven alone). Additionally, we conduct the first human evaluation to assess the utility of LLM-generated hypotheses in assisting human decision-making on two challenging tasks: deception detection and AI generated content detection. Our results show that human accuracy improves significantly by 7.44\% and 14.19\% on these tasks, respectively. These findings suggest that integrating literature-based and data-driven approaches provides a comprehensive and nuanced framework for hypothesis generation and could open new avenues for scientific inquiry.

cs.AI↗