large language models

Space-Folding Bounce Cosmology Without Ghost: A R^2-Corrected Non-Singular Origin of the Big Bang

The standard \LambdaCDM cosmological model inevitably encounters a spacetime singularity when extrapolated backward in time, where the scale factor a \to 0, density \rho \to \infty, and general relativity breaks down. While the Penrose-Hawking singularity theorems formalize this breakdown, they do not specify what physical mechanism replaces the singularity. Previous bounce models attempting to resolve this issue often rely on phantom scalar fields with negative kinetic energy, which introduce fatal ghost instabilities at the quantum level. In this work, we propose a ghost-free Space-Folding Bounce Model that eliminates the Big Bang singularity through a classical elastic rebound mechanism without invoking phantom fields or ad hoc quantum gravity inputs. The central hypothesis is that three-dimensional space possesses a foldable, elastic degree of freedom described by a canonical (positive-kinetic) scalar field \phi — the folding field — coupled to R^2 modified gravity. The action is constructed as S = \int d^4x\sqrt{-g}\left[\frac{M_{\rm Pl}^2}{2}\left(R + \frac{R^2}{6\rho_c}\right) + \frac{1}{2}\partial_\mu\phi\partial^\mu\phi - V_{\rm fold}(\phi)\right], where the saturating potential V_{\rm fold}(\phi) = V_0 + V_{\max}[1-(1+\phi/\phi_c)^{-\alpha}] encodes the finite compressibility of space. We demonstrate that: (i) the folding field has strictly positive energy density \rho_\phi = \frac{1}{2}\dot\phi^2 + V \ge 0, thereby evading ghost instabilities; (ii) the bounce occurs at finite a_{\min} > 0 driven by geometric corrections from R^2 terms, with H=0 and \ddot{a}/a > 0; (iii) the strong energy condition is violated by the effective total stress-energy at the bounce point; (iv) numerical integration of the coupled field equations with constraint-projected initial conditions confirms smooth solutions crossing the bounce with all physical quantities finite. The model reduces to standard \LambdaCDM in the low-folding limit \phi \to 0, with residual potential V_0 providing an effective cosmological constant. Observable predictions include a spectral index n_s \approx 0.964, tensor-to-scalar ratio r \approx 0.03, and local-type primordial non-Gaussianity f_{\rm NL}^{\rm local} \sim \mathcal{O}(1) — a characteristic fingerprint distinguishable from single-field slow-roll inflation. This model offers a complete, ghost-free, and observationally testable alternative to the singular Big Bang paradigm. AI Statement The theoretical framework, derivations, and numerical computations in this work were carried out by the authors. Large language model (LLM) tools were used to assist with text polishing, formula formatting, and code implementation. All results and conclusions have been independently verified and are the sole responsibility of the authors.

InventOR: A Prompt-Configured, No-Fine-Tuning LLM Workflow for Material-Level Inventory Control

We present InventOR, a prompt-configured large-language-model (LLM) workflow that generates material-level (r, Q) inventory policies from cutoff-bounded historical CSVs and inline parameter blocks through an enterprise LLM application, without project-specific model training or fine-tuning. Deterministic parsing and post-cutoff simulation evaluate each LLM artifact against an SAP-derived and an SAP-safety-lead-time-informed operations-research comparator on a common 346-pair cohort from a three-plant industrial dataset. All 365 eligible plant-material pairs yield scoreable, capacity-feasible artifacts after retry handling. Working-day simulation reports 77.52% demand-weighted aggregate fill and 97.94% mean material fill for the LLM-emitted arm. Runs 2 and 3 show repeated completion but policy-value variation. Run 1 lacks preserved deployed prompt and runtime metadata; the contribution is a transparent descriptive evaluation protocol with explicit boundaries and current-run provenance controls for prospective evaluation.

Vibration-Language Model for Fault Diagnosis with Numerically Reliable Evidence

Fault diagnosis methods for mechanical equipment should not only identify fault categories but also provide verifiable diagnostic evidence. Existing deep learning models usually output only labels or confidence scores, making it difficult to estab-lish an interpretable diagnostic reasoning process. Large language models have strong capabilities in evidence organization and explanation generation. However, a modality gap exists between vibration signals and discrete language tokens. In addi-tion, key numerical evidence in generated diagnostic reports may be inaccurate or hallucinated. To address these issues, this paper proposes a Vibration-Language Model (ViLM). The proposed method encodes angle-domain waveforms and order spectra into learnable vibration tokens and maps them into the embedding space of a large language model. With signal-description alignment and diagnostic instruction tuning, this architecture enables interpretable fault diagnosis based on vibration-informed language generation. Furthermore, a numerical evidence-constrained de-coding method is designed to embed the computation and backfilling of key numeri-cal evidence into the generation process. Experiments show that ViLM improves fault classification and evidence-supported explanation while maintaining high nu-merical reliability in generated diagnostic reports.

M-AIDA: Meta-Analysis Intelligent Data Assistant

Research software for international-business meta-analysis: semi-automated effect-size extraction from academic PDFs through a vendor-neutral large-language-model adapter, human-in-the-loop verification by the principal investigator, and an immutable data-lock workflow that exports a reproducible effect-size dataset for three-level meta-analytic regression. Built to support the P6 (meta-analysis) component of the first author's doctoral dissertation on the internationalization-performance relationship.

InventOR: A Prompt-Configured, No-Fine-Tuning LLM Workflow for Material-Level Inventory Control

We present InventOR, a prompt-configured large-language-model (LLM) workflow that generates material-level (r, Q) inventory policies from cutoff-bounded historical CSVs and inline parameter blocks through an enterprise LLM application, without project-specific model training or fine-tuning. Deterministic parsing and post-cutoff simulation evaluate each LLM artifact against an SAP-derived and an SAP-safety-lead-time-informed operations-research comparator on a common 346-pair cohort from a three-plant industrial dataset. All 365 eligible plant-material pairs yield scoreable, capacity-feasible artifacts after retry handling. Working-day simulation reports 77.52% demand-weighted aggregate fill and 97.94% mean material fill for the LLM-emitted arm. Runs 2 and 3 show repeated completion but policy-value variation. Run 1 lacks preserved deployed prompt and runtime metadata; the contribution is a transparent descriptive evaluation protocol with explicit boundaries and current-run provenance controls for prospective evaluation.

M-AIDA: Meta-Analysis Intelligent Data Assistant

Research software for international-business meta-analysis: semi-automated effect-size extraction from academic PDFs through a vendor-neutral large-language-model adapter, human-in-the-loop verification by the principal investigator, and an immutable data-lock workflow that exports a reproducible effect-size dataset for three-level meta-analytic regression. Built to support the P6 (meta-analysis) component of the first author's doctoral dissertation on the internationalization-performance relationship.

A Real-World Evaluation of Large Language Model–Generated Hospital Courses in Pediatrics

BACKGROUND Large language model (LLM)–generated hospital courses are increasingly integrated into electronic health records (EHRs), yet their accuracy and safety in pediatric populations remain poorly characterized. OBJECTIVE To evaluate the accuracy, text quality, and perceived potential harm of EHR-integrated and LLM-generated hospital courses in pediatric inpatient care during early clinical implementation. METHODS We conducted a descriptive evaluation from June 10 to August 8, 2025, at an academic freestanding children’s hospital using an Epic EHR with an integrated LLM tool (GPT-4o and GPT-4.1). Clinicians across multiple roles, including attending physicians, residents, and advanced practice providers, reviewed LLM-generated hospital courses for their own patients. Clinicians identified and categorized errors (hallucinations, inaccuracies, or omissions). They also rated text quality (comprehensiveness, conciseness, coherence) on a 5-point scale and perceived harm on an 8-point scale. RESULTS A total of 129 LLM-generated hospital courses were reviewed (median length of stay, 3 days; IQR, 2–7) by 50 involved clinicians. Hallucinations occurred in 21% (95% CI, 14%-29%) of the hospital courses, inaccuracies in 41% (53/129; 95% CI, 33%-50%), and omissions in 24% (31/129; 95% CI, 17%-32%). Overall, perceived harm ratings were low (median, 0; IQR, 0–1). Text quality ratings were high (median [IQR]: comprehensiveness, 4 [3–5]; conciseness, 4 [4–5]; coherence, 4 [4–5]) and comparable with prior literature. CONCLUSION In this pediatric evaluation of LLM-generated hospital courses reviewed by frontline clinicians, errors were common, but perceived potential harm was low, even assuming use without clinician correction. These findings support the use of LLM-generated hospital courses as starting drafts when paired with clinician review and institutional safeguards.

M-AIDA: Meta-Analysis Intelligent Data Assistant

Research software for international-business meta-analysis: semi-automated effect-size extraction from academic PDFs through a vendor-neutral large-language-model adapter, human-in-the-loop verification by the principal investigator, and an immutable data-lock workflow that exports a reproducible effect-size dataset for three-level meta-analytic regression. Built to support the P6 (meta-analysis) component of the first author's doctoral dissertation on the internationalization-performance relationship.

AI-Driven IoT (AIIOT) in Brain Health Study

Context and Justification The prevalence of neurodegenerative diseases around the world demands a paradigm change from reactive, episodic clinical diagnosis to ongoing, proactive neuro-monitoring. The possibility for early detection and individualized management is limited by the fact that traditional diagnostic techniques sometimes rely on subjective evaluation or costly, intrusive imaging. An unparalleled chance to identify, evaluate, and interpret the subtle, objective indicators of preclinical cognitive alterations is presented by the convergence of the Internet of Things (IoT) and sophisticated artificial intelligence (AI). Methods In order to generate a continuous, longitudinal stream of physiological and behavioural data, this study presents a novel, decentralized platform that makes use of a heterogeneous network of IoT devices, such as high-resolution wearables, smart home sensors, and non-contact physiological monitors (digital phenotyping). Large datasets pertaining to sleep architecture, gait variability, social interaction frequency, and speech hesitancy were analysed using deep learning techniques, particularly convolutional neural networks for anomaly detection in sensor data and long short-term memory (LSTM) networks for temporal pattern recognition. In order to forecast the start of moderate cognitive impairment months before conventional clinical criteria could be satisfied, the main goal was to train these AI models to recognize minute variations from each person’s unique baseline. Important Results (Hypothetical) With a 92% prediction accuracy, the AI-driven study was able to identify a multivariate biomarker profile associated with early cognitive deterioration. Importantly, the system was able to identify temporary changes in everyday activities, such as increased nocturnal wandering and entropy changes in spoken language, long before carers noticed them or could measure them using conventional paper-and-pencil exams. Real-time anomaly notifications made possible by the incorporation of edge computing enabled prompt triage and focused clinical evaluation. Conclusion The neuro-sensing grid underlines how important AI-powered IoT is to revolutionizing research on brain health. This technique provides a reliable, scalable, and non-invasive method for personalized neuro-surveillance by moving the locus of assessment from the clinic to the lived environment. This opens the door for truly preventive therapies against age-related cognitive decline.

A Survey on Table Mining with Large Language Models: Challenges, Advancements and Prospects

Table mining is a popular research field that involves complicated technologies, including information retrieval, data mining, visual and textual understanding and logical reasoning. With the emergence of Large Language Models (LLMs), the field has witnessed considerable advancements, presenting new paradigms for table understanding, extraction, and reasoning. In this survey, we conduct a comprehensive review of the literature on table mining with LLMs. We begin by introducing the fundamental overview of tabular data and possible challenges in LLM-based table mining. Specifically, we explore the challenges unique to this domain, such as heterogeneous table structures, contextual ambiguity, and domain-specific knowledge requirements. Then, we summarize representative tabular tasks in table preparation and mining, categorizing existing methods along dimensions including task scope, model architecture, and application scenarios. Next, we describe advanced LLM-based learning strategies in table mining, including foundation models and training-free methods. We further review studies of trustworthy LLM-based table mining and some domain-specific applications. Finally, we discuss prospects and future directions in the field of LLM-based table mining, including issues of generalization, interpretability, efficiency, etc . We hope this survey provides a comprehensive resource for researchers and practitioners, paving the way for further exploration. The repository is at: https://github.com/USTCAGI/Awesome-LLM-Table-Mining.

Evaluating Large Language Models for Triple Extraction from Chinese Historical Geographic Sources: A Case Study Using Sources from the China Historical GIS Project

Multimodal medical diagnosis: a mini review of LLM–vision fusion models in low-resource healthcare settings

Recent advances in large language models (LLMs) and vision transformers have enabled multimodal systems that integrate clinical text with medical imaging for diagnostic decision-making. While these systems show promising results on benchmark datasets in well-resourced research settings, their applicability in low-resource healthcare environments where diagnostic disparities are most severe remains limited and poorly understood. This mini review synthesizes key developments in LLM–vision fusion architectures from 2018 to 2026, with a focus on radiology-oriented visual question answering (VQA) and report generation systems viewed from a deployment perspective. Rather than comprehensively cataloguing multimodal medical AI, we synthesize the evolution of LLM–vision fusion architectures and discuss complementary deployment-enabling strategies, including parameter-efficient adaptation, post-training quantization, federated learning, and multilingual support, where they directly improve the feasibility of radiology AI in resource-constrained healthcare settings. Rather than focusing solely on performance benchmarks, we examine these approaches through a deployment-oriented lens, highlighting trade-offs between representational capacity, computational efficiency, interpretability, and memory footprint. We argue that current progress remains substantially shaped by model scaling and benchmark optimization, which often do not address the memory, connectivity, and annotation constraints of low-resource healthcare systems. While cross-modal transformer architectures provide strong representational alignment, their computational demands and reliance on large curated datasets limit real-world deployment. In contrast, emerging directions including parameter-efficient fine-tuning, post-training quantization, federated learning, and modular agent-based systems offer more tractable pathways toward clinical integration under hardware and data constraints. To bridge the gap between benchmark performance and clinical utility, we identify concrete challenges in data scarcity, multilingual coverage, and calibration, and propose a shift toward lightweight, interpretable, and hardware-aware multimodal AI. This perspective highlights the need to move beyond scaling-centric design toward models that can run on 4–8 GB VRAM, operate offline, and generalize across languages and imaging equipment.