arXiv subjects

Junyu Mao

Publications and source records attributed to Junyu Mao.

9 recordsLinked to original sources

arXiv2026

Automated Data Enrichment using Confidence-Aware Fine-Grained Debate among Open-Source LLMs for Mental Health and Online Safety

Real-world indicators play an important role in many Natural Language Processing (NLP) applications, such as life events for mental health analysis and risky behaviours for online safety, yet labelling such information is often costly and/or difficult due to its multi-label and dynamic nature. Large Language Models (LLMs) show promising potential for automated annotation, but the multi-label setting remains challenging. In this work, we propose a Confidence-Aware Fine-Grained Debate (CFD) framework that simulates human collaborative annotation using fine-grained communication to better support automated multi-label enrichment. We introduce two expert-annotated resources: life-event and symptom annotation for a mental health well-being dataset, and a new online safety sharenting dataset. Experiments show that CFD achieves the most robust enrichment performance across tasks, and that the benefit of fine-grained confidence is influenced by its quality and variability. We further evaluate training-free strategies for incorporating enrichment indicators into downstream tasks and show that automated enrichment consistently improves performance, by up to 9.9 Macro-F1 points, with the most effective integration format depending on how the indicator relates to the downstream objective.

cs.CL↗

arXiv2026

SCOUT: Semantic scene COverage via Uncertainty-guided Traversal

Robots that operate over extended periods should not merely visit space; they should progressively understand it. Yet most 3D scene graph pipelines treat perception as a post-processing stage over a fixed dataset, decoupling scene representation from the decisions that determine what is observed in the first place. We present SCOUT, an online semantic exploration framework that closes this loop by coupling active traversal with probabilistic scene graph construction. Given a prior 2D occupancy map and posed RGB-D observations, SCOUT incrementally builds an uncertainty-aware 3D scene graph whose nodes maintain fused geometry and posterior beliefs over open-vocabulary object labels, while edges encode structural relations such as on, inside, belong, and next to. These beliefs are fed back to an uncertainty-guided traversal planner, which selects viewpoints by balancing expected semantic certainty gain, geometric coverage gain, and travel cost. In this way, the robot revisits ambiguous objects when additional evidence matters and expands into unseen free space when the scene remains incomplete. The resulting system treats semantic scene completeness as an operational objective rather than a passive by-product of semantic mapping, moving toward autonomous agents that can patrol, update, and reason about evolving indoor environments with minimal human intervention.

cs.RO↗

arXiv2025

One Equation to Rule Them All -- Part I: Direct Data-Driven Cascade Stabilisation

In this article we present a framework for direct data-driven control for general problems involving interconnections of dynamical systems. We first develop a method to determine the solution of a Sylvester equation from data. Such solution is used to describe a subspace that plays a role in a large variety of problems. We then provide an error analysis of the impact that noise has on this solution. This is a crucial contribution because, thanks to the interconnection approach developed throughout the article, we are able to track how the noise propagates at each stage, and thereby provide bounds on the final designs. Among the many potential problems that can be solved with this framework, we focus on three representatives: cascade stabilisation, model order reduction, and output regulation. This manuscript studies the first problem, while the companion Part II addresses the other two. For each of these settings we show how the problems can be recast in our framework. In the context of cascade stabilisation, we consider the 2-cascade problem, the effect of noise through the cascade, as well as N-cascade case, and we demonstrate that our proposed method is data efficient. The proposed designs are illustrated by means of a numerical example.

eess.SY↗

arXiv2025

One Equation to Rule Them All -- Part II: Direct Data-Driven Reduction and Regulation

The Sylvester equation underpins a wide spectrum of control synthesis and systems analysis tools associated with cascade interconnections. In the preceding Part I [1] of this article, it was shown that such an equation can be reformulated using data, enabling the production of a collection of data-driven stabilisation procedures. In this second part of the article, we continue to develop the framework established in Part I to solve two important control-theoretic problems: model order reduction and output regulation. For the model order reduction problem we provide a solution from input-state measurements, from input-output measurements, and we study the effect of the noise. For the output regulation problem, we provide data-driven solutions for the static and dynamic feedback problem. The proposed designs are illustrated by means of examples.

eess.SY↗

arXiv2025

Data-Driven Model Reduction by Moment Matching for Linear and Nonlinear Parametric Systems

Theory and methods to obtain parametric reduced-order models by moment matching are presented. The definition of the parametric moment is introduced, and methods (model-based and data-driven) for the approximation of the parametric moment of linear and nonlinear parametric systems are proposed. These approximations are exploited to construct families of parametric reduced-order models that match the approximate parametric moment of the system to be reduced and preserve key system properties such as asymptotic stability and dissipativity. The use of the model reduction methods is illustrated by means of a parametric benchmark model for the linear case and a large-scale wind farm model for the nonlinear case. In the illustration, a comparison of the proposed approximation methods is drawn and their advantages/disadvantages are discussed.

eess.SY↗

arXiv2024

Model Order Reduction of Large-Scale Wind Farms: A Data-Driven Approach

This paper proposes a data-driven algorithm for model order reduction (MOR) of large-scale wind farms and studies the effects that the obtained reduced-order model (ROM) has when this is integrated into the power grid. With respect to standard MOR methods, the proposed algorithm has the advantages of having low computational complexity and not requiring any knowledge of the high order model. Using time-domain measurements, the obtained ROM achieves the moment matching conditions at selected interpolation points (frequencies). With respect to the state of the art, the method achieves the so-called two-sided moment matching, doubling the accuracy by doubling the interpolated points. The proposed algorithm is validated on a combined model of a 200-turbine wind farm (which is reduced) interconnected to the IEEE 14-bus system (which represents the unreduced study area) by comparing the full-order model and the reduced-order model in terms of their Bode plots, eigenvalues and the point of common coupling voltages in extensive fault scenarios of the integrated power system.

eess.SY↗

arXiv2024

Do prompt positions really matter?

Prompt-based models have gathered a lot of attention from researchers due to their remarkable advancements in the fields of zero-shot and few-shot learning. Developing an effective prompt template plays a critical role. However, prior studies have mainly focused on prompt vocabulary searching or embedding initialization within a predefined template with the prompt position fixed. In this empirical study, we conduct the most comprehensive analysis to date of prompt position for diverse Natural Language Processing (NLP) tasks. Our findings quantify the substantial impact prompt position has on model performance. We observe that the prompt positions used in prior studies are often sub-optimal, and this observation is consistent even in widely used instruction-tuned models. These findings suggest prompt position optimisation as a valuable research direction to augment prompt engineering methodologies and prompt position-aware instruction tuning as a potential way to build more robust models in the future.

cs.CL↗

arXiv2024

MELA: Multilingual Evaluation of Linguistic Acceptability

In this work, we present the largest benchmark to date on linguistic acceptability: Multilingual Evaluation of Linguistic Acceptability -- MELA, with 46K samples covering 10 languages from a diverse set of language families. We establish LLM baselines on this benchmark, and investigate cross-lingual transfer in acceptability judgements with XLM-R. In pursuit of multilingual interpretability, we conduct probing experiments with fine-tuned XLM-R to explore the process of syntax capability acquisition. Our results show that GPT-4o exhibits a strong multilingual ability, outperforming fine-tuned XLM-R, while open-source multilingual models lag behind by a noticeable gap. Cross-lingual transfer experiments show that transfer in acceptability judgment is non-trivial: 500 Icelandic fine-tuning examples lead to 23 MCC performance in a completely unrelated language -- Chinese. Results of our probing experiments indicate that training on MELA improves the performance of XLM-R on syntax-related tasks. Our data is available at https://github.com/sjtu-compling/MELA.

cs.CL↗

arXiv2022

Data-Driven Model Reduction by Two-Sided Moment Matching

In this brief paper, we propose a time-domain data-driven method for model order reduction by two-sided moment matching for linear systems. An algorithm that asymptotically approximates the matrix product $ΥΠ$ from time-domain samples of the so-called two-sided interconnection is provided. Exploiting this estimated matrix, we determine the unique reduced-order model of order $ν$, which asymptotically matches the moments at $2 ν$ distinct interpolation points. Furthermore, we discuss the impact that certain disturbances and data distortions may have on the algorithm. Finally, we illustrate the use of the proposed methodology by means of a benchmark model.

eess.SY↗