large language models

A review of bias detection and fairness auditing techniques in LLMs

The rise of large language models (LLMs) has sparked worries about inherent social biases and issues related to fairness. Earlier studies have investigated bias identification in word embeddings, interventions aimed at fairness in algorithms, and frameworks for auditing at the system level. Nonetheless, these methods remain disorganized, with variations in datasets, evaluation methods, and implementation processes. In this paper, we provide a thorough literature review to encapsulate prior research on bias identification and fairness auditing, categorizing the findings according to various stages of study. Additionally, we analyze the limitations in coverage and consistency of widely used benchmark datasets. To tackle these issues, we propose a unified pipeline for dataset integration and a modular framework for bias auditing. Recognized significant research gaps include the absence of intersectional bias modeling, a shortage of standardized evaluation metrics, and challenges in scalability for real-time auditing systems.

Adapting NDBIs for Autistic Children Who Use Echolalia: A Cascading Coaching Approach to Supporting Communication and Language Development

This pilot study examined a cascading coaching model to train graduate student clinicians and caregivers to use adapted Naturalistic Developmental Behavioral Intervention (NDBI) techniques with autistic preschoolers who use echolalia. Adaptations were designed to incorporate children’s echolalic speech within intervention activities and support changes in spoken language. Strategies emphasized responding to, expanding upon, and creating opportunities around children’s verbal productions during play. Two clinician–caregiver–child triads participated. Effects were evaluated using multiple-probe designs across adult participants and intervention contexts. Across four weeks, adults increased their use of targeted NDBI skills, including elicitation techniques (e.g., creating opportunities, wait time, prompting) and response techniques (responsive actions and language models), with moderate to very large effects. Child participants also demonstrated increases in targeted spoken-language communication responses across intervention contexts, with the strongest effects observed during student clinician-led interactions and more variable outcomes during caregiver-led interactions. Visual analysis indicated functional relations for most adult and child outcomes, although individual differences and variability were observed across participants. Findings provide preliminary evidence regarding the feasibility of incorporating echolalia-responsive adaptations within NDBI-based coaching programs for autistic children who use echolalia and their caregivers.

SciRep: A Ranking-Aware Representation Model for Scientific Text

Existing scientific text representation methods based on contrastive learning typically adopt a binary classification paradigm of positive and negative samples, which struggles to capture the complex, hierarchical semantic similarity relationships inherent in scientific texts. To address this, we propose SciRep, a novel two-stage ranking distillation framework. In the first stage, we distill knowledge from a large language model to a medium-scale representation model using generated ranking samples; in the second stage, a multi-teacher strategy further transfers fine-grained ranking capability to a lightweight model. Evaluated on a scientific literature semantic embedding benchmark comprising three tasks, SciRep outperforms the strongest baseline by 11.3% in terms of Average Rank and also achieves the highest Mean Reciprocal Rank scores across all three tasks. These results demonstrate that the proposed ranking-aware distillation mechanism significantly enhances scientific text representation quality while maintaining efficient inference, offering a more effective contrastive learning method for domain-specific retrieval tasks.

Evaluation of a cornea-specialized large language model for diagnostic and management accuracy in complex corneal cases

Retrieval Augmented Generation for Relative Distance-based Spatial Reasoning

Large Language Models (LLMs) can be applied to various real-world tasks, many of which require spatial reasoning, like pattern-based spatial search, Point of Interest (POI) recommendation, and itinerary or travel planning. However, on their own LLMs have been shown to lack reliable spatial reasoning capabilities, especially about distances. Further, success at these tasks requires the LLM to have been exposed to the relevant places and to successfully disambiguate their spatial interpretations, which is affected by the popularity of the place name. To address these problems, we develop a novel approach using spatial Retrieval Augmented Generation (RAG) to enable an LLM to use relevant distance information not explicitly learned during training when answering spatial questions that involve distance reasoning. Our method encodes a spatial graph of geodesic and road network distances between cities and towns, as well as a mapping to their alternative names, and compares a variety of techniques to provide the LLM with the appropriate spatial context from those knowledge stores. By incorporating relevant spatial information, our method significantly outperforms the baseline LLM on several tasks requiring distance-based spatial reasoning, including nearest-neighbor questions and distance analogy questions. We further show through experiments which graph storage and retrieval methods perform best, and which are brittle when applied to more challenging questions. Given the vast array of possible places an LLM could be asked to reason about, our work offers a first step towards providing a rudimentary distance-oriented spatial ‘world model’ to complement the vast linguistic knowledge held in LLMs.

Generative large language models in medicine: a scoping review of recent methodological advances

Comparative Analysis of Aesthetic Judgment Between Orthodontists, General Dentists, and Different Versions of Large Language Models Using the IOTN-AC Index

OpenAsset: A Pipeline for Open-World Asset Integration into Indoor Scene Synthesis

Text-driven 3D indoor scene synthesis has witnessed significant progress through Large Language Model (LLM)-based frameworks like ReSpace. However, current paradigms heavily rely on retrieving objects from pre-defined, static 3D asset libraries, which fundamentally constrains the diversity and personalization of generated scenes due to the closed-set nature of existing databases. Conversely, recent breakthroughs in promptable segmentation (SAM 3) and single-image 3D reconstruction (SAM 3D) have empowered the extraction of high-fidelity 3D geometry and texture from in-the-wild images. In this paper, we bridge the gap between text-driven scene layout generation and single-view object reconstruction by proposing OpenAsset. This novel framework converts single images of real-world objects into reusable, canonicalized 3D assets that seamlessly integrate into the scene synthesis workflow. Specifically, given a user concept prompt or target region, OpenAsset leverages SAM3 for precise instance isolation and SAM3D for geometry and texture recovery. To ensure compatibility with structured scene representations (SSR), we introduce an automated canonicalization module that normalizes the scale, orientation, and coordinate systems of the reconstructed meshes. By transforming “wild” visual percepts into standardized assets, OpenAsset effectively expands the controllable vocabulary of indoor scene synthesis beyond curated datasets, offering a practical pathway toward user-sourced open 3D scene generation that supports custom objects outside fixed predefined asset libraries. It is worth noting that the satisfactory performance of OpenAsset critically depends on effective segmentation and reconstruction results, which serve as essential prerequisites for our method.

A Knowledge-Enhanced Iterative Reasoning Framework for Accurate and Traceable Fault Diagnosis in Distributed Service Systems

Fault diagnosis in distributed systems is challenged by complex service dependencies, cascading anomaly propagation, and similar symptom patterns. This paper proposes a knowledge-enhanced iterative reasoning framework that integrates large language models (LLMs) with a numerical domain knowledge graph (KG). The KG encodes fault–symptom relations, anomaly directions, and training-derived mean and standard-deviation intervals. Structured prompting first generates candidate faults; interval verification then rejects numerically inconsistent candidates. For retained candidates, counterfactual reasoning constructs hierarchical causal chains, KG traversal refines missing or inconsistent links, and a deterministic evidence score supports acceptance, exclusion, early stopping, and fallback across at most five iterations. Under the common 68-case evaluation protocol for eight known single-root-cause faults in the controlled Redis-based testbed, the complete framework achieved 100.00% Accuracy, Macro-F1, and Balanced Accuracy with GPT-4o and GPT-5.2, compared with 91.18% accuracy for KG-only reasoning and 85.29–89.71% for Random Forest, XGBoost, and Transformer baselines. GPT-3.5 reached 98.53%, whereas LLaMA-3.1-8B reached 80.88%, showing that the incremental KG–LLM gain is backbone-dependent. Five GPT-4o repetitions and three GPT-5.2 repetitions yielded 100.00% ± 0.00, and all three metrics remained at 100.00% across the evaluated Z-score thresholds, iteration limits, and interval tolerances. The framework therefore provides highly accurate, stable, and traceable diagnoses within the evaluated Redis-based distributed-service protocol, while providing explicit intermediate reasoning and solution retrieval.

A multimodal benchmark for editable constraint preserving history based CAD modeling

Editable CAD generation requires access to procedural history rather than final geometry alone, yet widely used CAD datasets typically omit explicit geometric constraints, native parametric assets, or rich semantic labels. We introduce EPICCAD, a multimodal benchmark for history-based CAD modeling that aligns compact constraint-aware modeling sequences with multi-view renderings, STEP B-reps, native parametric files, and textual annotations. Importantly, EPICCAD flattens only the redundant sketch-loop nesting inside individual profiles; it retains the procedural feature history and reconstructs loop topology at execution time, so geometric relations are made explicit through constraints rather than discarded. EPICCAD combines 152,360 academic samples with 8141 industrial Siemens NX parts, yielding 160,501 models that better reflect real design complexity. We further present AM $$_\text {EPICCAD}$$ , an annotation pipeline that parses loop structure, inter-part relations, and operation histories from CAD sequences and then uses a large language model to generate process-aware, geometry-aware, and function-aware descriptions. Experiments show that EPICCAD’s flat sequence representation reduces token length while preserving geometric fidelity, explicit constraints substantially improve editability, and both the annotation module and industrial split strengthen text-to-CAD generation. EPICCAD therefore provides a practical foundation for AI systems that must generate, interpret, and revise executable CAD models.

Large Language Models Versus Multidisciplinary Tumor Board Decisions in Thyroid Cancer

ABSTRACT Objectives Large language models (LLMs) are increasingly proposed as clinical decision‐support tools; however, their agreement with real‐world multidisciplinary tumor board (MDT) decisions remains insufficiently investigated in thyroid oncology. To evaluate the concordance between treatment recommendations generated by ChatGPT 5.2 and Gemini 3.0 and decisions made by a tertiary multidisciplinary thyroid tumor board. Methods This study included 59 consecutive patients discussed at a tertiary MDT between August and December 2025. Anonymized clinical data, including demographics, ultrasonographic findings, and Bethesda cytology, were provided to both LLMs using standardized structured prompts. MDT decisions were defined as the reference standard. Agreement was assessed using exact concordance rates and Cohen's kappa ( κ ) statistics with 95% confidence intervals. Results ChatGPT 5.2 achieved a concordance rate of 71.2% (42/59), demonstrating substantial agreement ( κ = 0.623; 95% CI 0.459–0.771). Gemini 3.0 showed a concordance rate of 64.4% (38/59), reflecting moderate agreement ( κ = 0.527; 95% CI 0.359–0.684). Discordance increased in complex scenarios involving lateral neck dissection, radioactive iodine therapy, and active surveillance. Conclusions While LLMs demonstrate promising concordance in standardized thyroid cancer management, they are best positioned as supportive decision aids—such as in MDT preparation and workflow streamlining—rather than replacements for expert multidisciplinary evaluation, particularly in complex clinical scenarios. Level of Evidence 3.

A Project-based Learning Activity Involving Hybrid Intelligence in the Undergraduate Engineering Class

Artificial intelligence (AI) is increasingly embedded in academic practice, creating opportunities to support learning and improve educational outcomes. We implemented a project-based learning activity in an undergraduate engineering course in which students used meteorological data to construct a temperature time series and assess local climate change at a selected location from the national weather station network. They transformed raw datasets, organized files in spreadsheets, and automated routine calculations by prompting a large language model (LLM) to generate executable code for a collaborative programming platform. As a structured form of human–AI collaboration, the activity contributes to discussions of hybrid intelligence in education by allowing students to focus on higher-level cognitive tasks.