arXiv subjects

Yandong Sun

Publications and source records attributed to Yandong Sun.

12 recordsLinked to original sources

arXiv2026

SPARROW: Scalable Taxonomy Induction via Structure-Preserving Partitioning and Constraint-Guided Merging

Taxonomy induction aims to organize concept sets into coherent hierarchical structures. Recent LLM-based methods can induce taxonomies directly from flat term lists, avoiding the need for corpora, but degrade sharply as concept sets scale up. We argue that this degradation stems not only from context length limitations, but also from structural failures in hierarchical reasoning. To address this, we adopt a divide-and-merge paradigm that partitions concepts into smaller subsets, induces local taxonomies, and merges them into a global hierarchy. However, we identify two structural failure modes inherent to this paradigm: Structural Fragmentation, where partitioning weakens local hierarchical signals, and Parent Displacement, where locally plausible relations are misplaced in the global hierarchy. To address both, we propose SPARROW, a scalable taxonomy induction framework that combines structure-preserving spectral partitioning to retain hierarchical connectivity within each block, and constraint-guided incremental fusion that treats block-level relations as structural constraints rather than ground truth for global placement. Experiments on large-scale benchmarks show that SPARROW consistently achieves the strongest global structural quality across backbones. The code is available at https://github.com/rebeccazyr/SPARROW.

cs.CL↗

arXiv2026

Emergency Preemption Without Online Exploration: A Decision Transformer Approach

Emergency vehicle (EV) response time is a critical determinant of survival outcomes, yet deployed signal preemption strategies remain reactive and uncontrollable. We propose a return-conditioned framework for emergency corridor optimization based on the Decision Transformer (DT). By casting corridor optimization as offline, return-conditioned sequence modeling, our approach (1) eliminates online environment interaction during policy learning, (2) enables dispatch-level urgency control through a single target-return scalar, and (3) extends to multi-agent settings via a Multi-Agent Decision Transformer (MADT) with graph attention for spatial coordination. On the LightSim simulator, DT reduces average EV travel time by 37.7% relative to fixed-timing preemption on a 4x4 grid (88.6 s vs. 142.3 s), achieving the lowest civilian delay (11.3 s/veh) and fewest EV stops (1.2) among all methods, including online RL baselines that require environment interaction. MADT further improves on larger grids, overtaking DT with 45.2% reduction on 8x8 via graph-attention coordination. Return conditioning produces a smooth dispatch interface: varying the target return from 100 to -400 trades EV travel time (72.4-138.2 s) against civilian delay (16.8-5.4 s/veh), requiring no retraining. A Constrained DT extension adds explicit civilian disruption budgets as a second control knob.

cs.LG↗

arXiv2026

Asking the Right Questions: Ontology-Grounded Interpretable Embeddings for Biomedical Text

While dense biomedical embeddings achieve strong performance, their opaque dimensions limit transparency in biomedical NLP. Recent question-based interpretable embeddings represent text through binary answers to natural-language questions, but existing approaches rely primarily on corpus-driven signals, often capturing topical or stylistic differences rather than fine-grained biomedical distinctions. We propose QIME, an ontology-grounded framework for interpretable biomedical text embeddings in which each dimension corresponds to an explicit biomedical-domain yes/no question. QIME leverages a biomedical ontology to guide contrastive question generation from semantic clusters, producing atomic, domain-grounded questions. It constructs embeddings via similarity-based semantic activation with MMR-based diversity-aware dimension selection, yielding sparse representations efficiently. Experiments on biomedical clustering, STS and retrieval benchmarks show that QIME consistently outperforms prior interpretable embedding methods and substantially narrows the gap to strong black-box biomedical encoders. It also imposes substantially lower cognitive burden than existing methods, providing concise and domain-specific interpretations. The code is available at https://github.com/L1nzh/QIME.

cs.CL↗

arXiv2026

Spatiotemporal Decision Transformer for Traffic Coordination

Traffic signal control is a critical challenge in urban transportation, requiring coordination among multiple intersections to optimize network-wide traffic flow. While reinforcement learning has shown promise for adaptive signal control, existing methods struggle with multi-agent coordination and sample efficiency. We introduce MADT (Multi-Agent Decision Transformer), a novel approach that reformulates multi-agent traffic signal control as a sequence modeling problem. MADT extends the Decision Transformer paradigm to multi-agent settings by incorporating: (1) a graph attention mechanism for modeling spatial dependencies between intersections, (2) a|temporal transformer encoder for capturing traffic dynamics, and (3) return-to-go conditioning for target performance specification. Our approach enables offline learning from historical traffic data, with architecture design that facilitates potential online fine-tuning. Experiments on synthetic grid networks and real-world traffic scenarios demonstrate that MADT achieves state-of-the-art performance, reducing average travel time by 5-6% compared to the strongest baseline while exhibiting superior coordination among adjacent intersections.

cs.LG↗

arXiv2026

The End of Reward Engineering: How LLMs Are Redefining Multi-Agent Coordination

Reward engineering, the manual specification of reward functions to induce desired agent behavior, remains a fundamental challenge in multi-agent reinforcement learning. This difficulty is amplified by credit assignment ambiguity, environmental non-stationarity, and the combinatorial growth of interaction complexity. We argue that recent advances in large language models (LLMs) point toward a shift from hand-crafted numerical rewards to language-based objective specifications. Prior work has shown that LLMs can synthesize reward functions directly from natural language descriptions (e.g., EUREKA) and adapt reward formulations online with minimal human intervention (e.g., CARD). In parallel, the emerging paradigm of Reinforcement Learning from Verifiable Rewards (RLVR) provides empirical evidence that language-mediated supervision can serve as a viable alternative to traditional reward engineering. We conceptualize this transition along three dimensions: semantic reward specification, dynamic reward adaptation, and improved alignment with human intent, while noting open challenges related to computational overhead, robustness to hallucination, and scalability to large multi-agent systems. We conclude by outlining a research direction in which coordination arises from shared semantic representations rather than explicitly engineered numerical signals.

cs.AI↗

arXiv2025

One Swallow Does Not Make a Summer: Understanding Semantic Structures in Embedding Spaces

Embedding spaces are fundamental to modern AI, translating raw data into high-dimensional vectors that encode rich semantic relationships. Yet, their internal structures remain opaque, with existing approaches often sacrificing semantic coherence for structural regularity or incurring high computational overhead to improve interpretability. To address these challenges, we introduce the Semantic Field Subspace (SFS), a geometry-preserving, context-aware representation that captures local semantic neighborhoods within the embedding space. We also propose SAFARI (SemAntic Field subspAce deteRmInation), an unsupervised, modality-agnostic algorithm that uncovers hierarchical semantic structures using a novel metric called Semantic Shift, which quantifies how semantics evolve as SFSes evolve. To ensure scalability, we develop an efficient approximation of Semantic Shift that replaces costly SVD computations, achieving a 15~30x speedup with average errors below 0.01. Extensive evaluations across six real-world text and image datasets show that SFSes outperform standard classifiers not only in classification but also in nuanced tasks such as political bias detection, while SAFARI consistently reveals interpretable and generalizable semantic hierarchies. This work presents a unified framework for structuring, analyzing, and scaling semantic understanding in embedding spaces.

cs.AI↗

arXiv2025

Don't Reinvent the Wheel: Efficient Instruction-Following Text Embedding based on Guided Space Transformation

In this work, we investigate an important task named instruction-following text embedding, which generates dynamic text embeddings that adapt to user instructions, highlighting specific attributes of text. Despite recent advancements, existing approaches suffer from significant computational overhead, as they require re-encoding the entire corpus for each new instruction. To address this challenge, we propose GSTransform, a novel instruction-following text embedding framework based on Guided Space Transformation. Our key observation is that instruction-relevant information is inherently encoded in generic embeddings but remains underutilized. Instead of repeatedly encoding the corpus for each instruction, GSTransform is a lightweight transformation mechanism that adapts pre-computed embeddings in real time to align with user instructions, guided by a small amount of text data with instruction-focused label annotation. We conduct extensive experiments on three instruction-awareness downstream tasks across nine real-world datasets, demonstrating that GSTransform improves instruction-following text embedding quality over state-of-the-art methods while achieving dramatic speedups of 6~300x in real-time processing on large-scale datasets. The source code is available at https://github.com/YingchaojieFeng/GSTransform.

cs.CL↗

arXiv2023

Van der Waals semiconductor InSe plastifies by phase transformation

Inorganic semiconductor materials are integral to various modern technologies, yet their brittleness and limited deformability/processability pose a significant challenge in the development of flexible, wearable, and miniaturized electronics. The recent discovery of room-temperature plasticity in a few inorganic semiconductors offers a promising pathway to address this challenge, but the deformation mechanisms of these materials remain unclear. Here, we investigate the deformation of InSe, a two-dimensional (2D) van der Waals (vdW) semiconductor with substantial plasticity. By developing a machine-learned deep potential, we perform atomistic simulations that capture the deformation features of hexagonal InSe upon out-of-plane compression. Surprisingly, we discover that InSe plastifies through a so-far unrecognized martensitic phase transformation; that is, the layered hexagonal structure is converted to a tetragonal lattice with specific orientation relationship. This observation is corroborated by high-resolution experimental observations and theory. It suggests a change of paradigm, where the design of new plastically-deformable inorganic semiconductors should focus on compositions and structures that favor phase transformations rather than traditional dislocation slip.

cond-mat.mtrl-sci↗

arXiv2021

Deep Learning Inter-atomic Potential for Thermal and Phonon Behaviour of Silicon Carbide with Quantum Accuracy

Silicon carbide (SiC) is an essential material for next generation semiconductors and components for nuclear plants. It's applications are strongly dependent on its thermal conductivity, which is highly sensitive to microstructures. Molecular dynamics (MD) simulation is the most used methods to address thermal transportation mechanisms in devices or microstructures of nano-meters. However, the implementation of MD is limited in SiC because of lacking accurate inter-atomic potentials. In this work, using the Deep Potential (DP) methodology, we developed two inter-atomic potentials (DP-IAPs) for SiC based on two adaptively generated datasets within the density functional approximations at the local density and the generalized gradient levels. These two DP-IAPs manifest their speed with quantum accuracy in lattice dynamics simulations as well as scattering rate analysis of phonon transportation. Combining with molecular dynamics simulations, the thermal transport and mechanical properties were systematically investigated. The presented methodology and the inter-atomic potentials pave the way for a systematic approach to model heat transport in SiC related devices using multiscale modelling.

cond-mat.mtrl-sci↗

arXiv2021

Phonon Scattering in the Complex Strain Field of a Dislocation

Strain engineering is critical to the performance enhancement of electronic and thermoelectric devices because of its influence on the material thermal conductivity. However, current experiments cannot probe the detailed physics of the phonon-strain interaction due to the complex, inhomogeneous, and long-distance features of the strain field in real materials. Dislocations provide us with an excellent model to investigate these inhomogeneous strain fields. In this study, non-equilibrium molecular dynamics simulations were used to study the lattice thermal conductivity of PbTe under different strain status tuned by dislocation densities. The extended 1D McKelvey-Shockley flux method was used to analyze the frequency dependence of phonon scattering in the inhomogeneously strained regions of dislocations. A spatially resolved phonon dislocation scattering process was shown, where the unequal strain in different regions affected the magnitude and frequency-dependence of the scattering rate. Our study not only advances the knowledge of strain scattering of phonon propagation but offers fundamental guidance on optimizing thermal management by structure design.

cond-mat.mtrl-sci↗

arXiv2021

Probing the Phonon Mean Free Paths in Dislocation Core by Molecular Dynamics Simulation

Thermal management is extremely important for designing high-performance devices. The lattice thermal conductivity of materials is strongly dependent on the structural defects at different length scales, particularly point defects like vacancies, line defects like dislocations, and planar defects such as grain boundaries. Traditionally, the McKelvey-Shockley phonon Boltzmann's transport equation (BTE) method combined with molecular dynamics simulations has been widely used to evaluate the phonon mean free paths (MFPs) in defective systems. However, this method can only provide the aggregate MFPs of the whole sample. It is, therefore, challenging to extract the MFPs in the different regions with different thermal properties. In this study, the 1D McKelvey-Shockley phonon BTE method was extended to model inhomogeneous materials, where the effect of defects on the phonon MFPs is explicitly obtained. Then, the method was used to study the phonon interactions with the core structure of an edge dislocation. The phonon MFPs in the dislocation core were obtained and consistent with the analytical model such that high frequency phonons are likely to be scattered in this area. This method not only advances the knowledge of phonon-dislocation scattering but also shows the potential to investigate phonon transport behaviors in more complicated materials.

cond-mat.mtrl-sci↗

arXiv2019

Strong phonon localization in PbTe with dislocations and large deviation to Matthiessen's rule

Dislocations can greatly enhance the figure of merit of thermoelectric materials by prominently reducing thermal conductivity. However, the evolution of phonon modes with different energies when they propagate through a single dislocation is unknown. Here we perform non-equilibrium molecular dynamics simulation to study phonon transport in PbTe crystal with dislocations by excluding boundary scattering and strain coupling effect. The frequency-dependent heat flux, phonon mode analysis, and frequency-dependent phonon mean free paths (MFPs) are presented. The thermal conductivity of PbTe with dislocation density on the order of 10^15m^-2 is decreased by 62%. We provide solid evidence of strong localization of phonon modes in dislocation sample. Moreover, by comparing the frequency-dependent phonon MFPs between atomistic modeling and traditional theory, it is found that the conventional theories are inadequate to describe the phonon behavior throughout the full phonon spectrum, and large deviation to the well-known semi-classical Matthiessen's rule is observed. These results provide insightful guidance for the development of PbTe based thermoelectrics and shed light on new routes for enhancing the performance of existing thermoelectrics by incorporating dislocations.

physics.app-ph↗