arXiv subjects

Shunshun Liu

Publications and source records attributed to Shunshun Liu.

6 recordsLinked to original sources

arXiv2026

Are LLMs Ready for Neural-integrated Mechanistic Modeling? A Benchmark and Agentic Framework

Large language models (LLMs) have shown promise in constructing mechanistic models from data. However, existing evaluations largely focus on simplified settings and fail to capture the complexity of real-world scientific modeling. In practice, such modeling often involves neural-integrated formulations, where a mechanistic model component and a neural network component are jointly constructed, leading to a significantly more complex search space. Motivated by this gap, we introduce the Neural-Integrated Mechanistic Modeling (NIMM) benchmark, which evaluates LLM-generated neural-integrated mechanistic models across three scientific domains. Experiments on NIMM reveal that existing LLM-based approaches struggle to effectively explore this complex space, resulting in limited search stability and solution quality. To address this challenge, we propose NIMMGen, a tree-guided agentic framework that enables diversified exploration via branch-level search and improves solutions through atomic model refinement. Extensive experiments demonstrate that NIMMGen achieves state-of-the-art performance on NIMM, significantly improving search stability and solution quality.

cs.LG↗

arXiv2025

Expert-Grounded Automatic Prompt Engineering for Extracting Lattice Constants of High-Entropy Alloys from Scientific Publications using Large Language Models

Large language models (LLMs) have shown promise for scientific data extraction from publications, but rely on manual prompt refinement. We present an expert-grounded automatic prompt optimization framework that enhances LLM entity extraction reliability. Using high-entropy alloy lattice constant extraction as a testbed, we optimized prompts for Claude 3.5 Sonnet through feedback cycles on seven expert-annotated publications. Despite a modest optimization budget, recall improved from 0.27 to > 0.9, demonstrating that a small, expert-curated dataset can yield significant improvements. The approach was applied to extract lattice constants from 2,267 publications, yielding data for 1,861 compositions. The optimized prompt transferred effectively to newer models: Claude 4.5 Sonnet, GPT-5, and Gemini 2.5 Flash. Analysis revealed three categories of LLM mistakes: contextual hallucination, semantic misinterpretation, and unit conversion errors, emphasizing the need for validation protocols. These results establish feedback-guided prompt optimization as a low-cost, transferable methodology for reliable scientific data extraction, providing a scalable pathway for complex LLM-assisted research tasks.

cs.DL↗

arXiv2025

Type-1.5 SNSPD: Interacting vortex theory of two bandgap superconducting single photon detectors

Photon detectors based on type-2 superconductors have found widespread applications from on-chip quantum computing to quantum remote sensing. Here, we develop the theory for a new class of type-1.5 superconducting nanowire single photon detectors (SNSPDs) based on two bandgap superconductors with high transition temperatures such as MgB2 (Tc ~38.6K). We show that vortex-vortex interactions in two component condensates lead to a unique operating regime where single photons can seed multiple vortices within a hotspot. We also show that dark counts are suppressed in the type-1.5 regime compared to the widely studied type-2 SNSPDs. Our work opens the door for exploring the unique vortex physics of two-gap superconductors for quantum device applications.

cond-mat.supr-con↗

arXiv2025

A Cognitive-Mechanistic Human Reliability Analysis Framework: A Nuclear Power Plant Case Study

Traditional human reliability analysis (HRA) methods, such as IDHEAS-ECA, rely on expert judgment and empirical rules that often overlook the cognitive underpinnings of human error. Moreover, conducting human-in-the-loop experiments for advanced nuclear power plants is increasingly impractical due to novel interfaces and limited operational data. This study proposes a cognitive-mechanistic framework (COGMIF) that enhances the IDHEAS-ECA methodology by integrating an ACT-R-based human digital twin (HDT) with TimeGAN-augmented simulation. The ACT-R model simulates operator cognition, including memory retrieval, goal-directed procedural reasoning, and perceptual-motor execution, under high-fidelity scenarios derived from a high-temperature gas-cooled reactor (HTGR) simulator. To overcome the resource constraints of large-scale cognitive modeling, TimeGAN is trained on ACT-R-generated time-series data to produce high-fidelity synthetic operator behavior datasets. These simulations are then used to drive IDHEAS-ECA assessments, enabling scalable, mechanism-informed estimation of human error probabilities (HEPs). Comparative analyses with SPAR-H and sensitivity assessments demonstrate the robustness and practical advantages of the proposed COGMIF. Finally, procedural features are mapped onto a Bayesian network to quantify the influence of contributing factors, revealing key drivers of operational risk. This work offers a credible and computationally efficient pathway to integrate cognitive theory into industrial HRA practices.

cs.AI↗

arXiv2023

Observation of helimagnetism in the candidate ferroelectric CrI$_2$

CrI$_{2}$ is a quasi-one dimensional (1D) van der Waals (vdW) system that exhibits helimagnetism that propagates along the ribbons. This was determined from neutron time-of-flight diffraction measurements. Below $T_N=17$ K, a screw-like helimagnetic order develops with an incommensurate wavevector of $\mathbf{q} \approx (0.2492,0,0)$ at 8 K. Using density functional theory (DFT)$+U$ calculations, the $J_{1}$-$J_{2}$ model was leveraged to describe the helimagnetism, where $J_{1} (> 0)$ and $J_2 (< 0)$ correspond, respectively, to a ferromagnetic nearest neighbor (NN) and antiferromagnetic next-nearest neighbor (NNN) intrachain interaction. The DFT$+U$ calculations predict that bulk CrI$_2$ in the orthorhombic $Cmc2_1$ crystal structure satisfies the $|J_2| > |J_1|/4$ condition, which favors formation of helimagnetic order.

cond-mat.mtrl-sci↗

arXiv2022

Integrating Machine Learning with Mechanistic Models for Predicting the Yield Strength of High Entropy Alloys

Accelerating the design of materials with targeted properties is one of the key materials informatics tasks. The most common approach takes a data-driven motivation, where the underlying knowledge is incorporated in the form of domain-inspired input features. Machine learning (ML) models are then built to establish the input-output relationships. An alternative approach involves leveraging mechanistic models, where the domain knowledge is incorporated in a predefined functional form. These mechanistic models are meticulously formulated through observations to validate specific hypotheses, and incorporate elements of causality missing from data-driven ML approaches. In this work, we demonstrate a computational approach that integrates mechanistic models with phenomenological and ML models to rapidly predict the temperature-dependent yield strength of high entropy alloys (HEAs) that form in the single-phase face-centered cubic (FCC) structure. Our main contribution is in establishing a quantitative relationship between the HEA compositions and temperature-dependent elastic constants. This allows us to improve the treatment of elastic constant mismatch to the solid solution strengthening effect in the mechanistic model, which is important for reliable prediction of yield strength as a function of temperature in single-phase FCC-based HEAs. We accomplish this by combining Bayesian inference with ensemble ML methods. The outcome is a probability distribution of elastic constants which, when propagated through the mechanistic model, yields a prediction of temperature-dependent yield strength, along with the uncertainties. The predicted yield strength shows good agreement with published experimental data, giving us confidence in applying the developed approach for the rapid search of novel FCC-based HEAs with excellent yield strength at various temperatures.

cond-mat.mtrl-sci↗