SolvedByAI

Queue

628 cases were harvested and are backed by a real paper, but they fail the test the index publishes on. Each one is listed here with the reason, so the exclusion can be checked rather than trusted.

All 628 assisted 358 not a result 183 unsupported 87
Paper Year Cited Why it is held
Can CNN‐based species classification generalise across variation in habitat... Methods in Ecology and Evolution 2022 31 assisted The paper is an evaluation of CNN classification performance and generalisability within a human-run ecological monitoring workflow, reporting accuracy metrics on an existing task rather than contributing the classification capability itself as a solved result. “Overall accuracy (72.8%) was improved by excluding the rarest species and by adding extra training images (76.3% and 82.8%, respectively).”
Human disease prediction from microbiome data by multiple feature fusion and... iScience 2022 30 assisted This is a methods paper reporting a human-developed framework's competitive benchmark accuracy on existing microbiome disease-prediction tasks, where the deep learning model is an evaluated component demonstrating a method rather than an agent that itself solved the prediction problem. “Experimental results indicate that MetaDR achieves competitive prediction performance with a reduction in running time”
Validation of a Deep Learning Chest X-ray Interpretation Model: Integrating... Diagnostics 2023 29 assisted This is a human-run validation study whose contribution is radiologists' graded evaluation of two models' reading quality on an existing task, with the AI explicitly framed as an assistive technology rather than the agent that solved the problem. “two radiologists evaluated the readings provided by KARA-CXR and ChatGPT”
A Smart Contract Vulnerability Detection Method Based on Multimodal Feature... Mathematics 2023 29 assisted The paper proposes a deep-learning detection method and reports its classification accuracies on the existing, well-established task of smart contract vulnerability detection, making the model an evaluated detection tool rather than an AI agent that produced a concrete solved result such as newly discovered vulnerabilities or secured contracts. “Experimental results demonstrate that the proposed method attains detection accuracies of 85.73% for reentrancy vulnerabilities, 85.41% for timestamp...”
A hybrid learning strategy for structural damage detection Structural Health Monitoring 2020 29 assisted The paper proposes a hybrid classifier methodology and reports its 'promising performance' on the existing task of damage detection, with humans running the simulations, lab/in-situ tests, and evaluation, making the AI a component in a human-driven validation workflow rather than the agent whose produced result is the paper's contribution. “Results achieved so far attest a promising performance of the hybrid classifier.”
Development and Validation of a Quick Sepsis-Related Organ Failure Assessmen... Journal of Clinical Medicine 2020 29 assisted The paper's contribution is a retrospectively validated clinical decision-support tool that reports AUROC on an existing prediction task (mortality risk scoring, where qSOFA is the established baseline), built and evaluated within a human-run research workflow with output intended to inform clinicians rather than an AI performing a task end to end. “the AUROC of the qSOFA-based machine-learning model (0.86 [95% CI 0.85–0.87]) for three-day mortality was higher than that of the qSOFA scores (0.78 [...”
Machine learning based charge mobility prediction for organic semiconductors The Journal of Chemical Physics 2023 26 assisted The ML models serve as an accelerator for quantum-chemical transfer integral calculations, feeding a human-run charge-transport pipeline whose mobility results are validated against brute-force QC rather than being produced end-to-end by the AI. “We apply these models to studying charge transport in organic crystals with dynamic disorders at 300 K and obtain the charge mobility and anisotropy i...”
Novel machine learning framework for thermal conductivity prediction by crys... SmartMat 2021 25 assisted The abstract's contribution is a human-designed ML framework whose benchmark performance on the existing task of lattice thermal conductivity prediction is reported, so the AI's predictions serve to validate a methodology rather than constituting an AI-produced scientific result. “This machine learning algorithms framework showed superior learning and generalization ability.”
Design of a large language model for improving customer service in telecom o... Electronics Letters 2024 24 assisted The LLM+LangChain system functions as a knowledge-recommendation component within human-operated customer support workflows (its outputs are accepted or rejected by human agents), rather than autonomously delivering the end-to-end service result. “The implementation of this LLM‐based framework within telecom customer service systems has significantly sharpened the precision of knowledge recommen...”
Application of machine learning in breast cancer survival prediction using a... Scientific Reports 2024 23 assisted The paper reports comparative accuracy of trained models (108 models, best 85.56%) on an existing clinical prediction task whose output is explicitly meant to inform human care planning, making the AI a predictive component in a human clinical workflow rather than the agent that produced the paper's primary result. “According to external validation, the DNN model trained with the Shiraz dataset, considering all features, exhibited the highest accuracy (85.56%).”
Examining Deep Learning Architectures for Crime Classification and Predictio... Forecasting 2021 20 assisted The paper is primarily a comparative benchmark study evaluating deep learning configurations against existing methods and providing recommendations for configuration, rather than presenting the AI's crime prediction as a novel solved result. “a comparative study of 10 state-of-the-art methods against 3 different deep learning configurations is conducted”
Use of responsible artificial intelligence to predict health insurance claim... Exploration of Digital Health Technologies 2024 19 assisted This is a model-comparison study reporting prediction accuracy on an existing task, where the AI outputs are meant to inform human decisions on insurance policy formulation and pricing rather than the AI itself performing the end-to-end task. “The findings demonstrate that the XGBoost and RF models outperformed the other algorithms, displaying the highest R-squared values of 79% and 77% and...”
The future of cosmological likelihood-based inference: accelerated high-dime... The Open Journal of Astrophysics 2024 19 assisted The ML components (emulation, learned harmonic mean) serve as acceleration tools inside a human-designed Bayesian inference pipeline, demonstrated on simulated data to reproduce results already obtainable by traditional methods, rather than the AI itself producing a novel result. “Specifically, we combine (i) emulation, where a machine learning model is trained to mimic cosmological observables, e.g. CosmoPower-JAX; (ii) differe...”
A novel procedure for cable damage identification of cable-stayed bridge usi... Structural Health Monitoring 2024 18 assisted The paper's contribution is a human-engineered procedure (PSO-based model updating, FE simulation, hyperparameter tuning) in which CatBoost serves as one predictive component, validated by accuracy comparison against other ML methods on a benchmark bridge rather than the AI itself producing a novel damage-identification result. “a procedure is proposed to identify the cable damage in the cable-stayed bridges using the particle swarm optimization (PSO) and the categorical gradi...”
Embedding the weather prediction errors (WPE) into the photovoltaic (PV) for... Journal of Forecasting 2024 18 assisted The paper's stated novelty is human-engineered features and adjustment coefficients wrapped around standard ML models to improve accuracy metrics on an existing PV-forecasting task, making the AI a prediction component inside a human-designed and human-completed method rather than the agent that itself constitutes the result. “The novelty of the research we conduct consists in proposing several features and determining the coefficients to adjust the PVF based on WPE.”
Improving Credit Risk Assessment through Deep Learning-based Consumer Loan D... International Journal of Finance & Banking Studies (2147-4486) 2023 14 assisted The deep learning model only produces default-likelihood predictions that inform banks' human-run credit review workflow, and the abstract merely reports test-set accuracy on an existing prediction task rather than the AI performing credit risk assessment end to end. “The model was evaluated on a test set of 250 records and achieved a test set accuracy of 95.2%, correctly predicting the default state of 238 out of 2...”
Machine Learning Guided Discovery of Non‐Linear Optical Materials Advanced Theory and Simulations 2024 14 assisted The ML models serve as a screening component that proposes candidates, which are then confirmed by human-run DFT calculations, making this a prediction-inside-a-pipeline rather than an AI-produced discovery. “These findings are validated using density functional theory(DFT) calculations.”
SymbolNet: neural symbolic regression with adaptive dynamic pruning for comp... Machine Learning Science and Technology 2025 13 assisted The paper's contribution is a human-designed method/framework (SymbolNet with adaptive sparsity regularization) whose evidence is effectiveness/accuracy demonstrations on existing benchmark tasks, rather than an AI-produced result that itself constitutes the contribution. “we demonstrate the effectiveness of our model on the LHC jet tagging task (16 inputs), MNIST (784 inputs), and SVHN (3072 inputs)”
Phase Transformation Temperature Prediction in Steels via Machine Learning Materials 2024 13 assisted The LightGBM model is used as a tool within a human-driven study of influencing factors and mechanisms (feature engineering, SHAP/PDP interpretation, discussion of alloying effects), rather than the AI itself producing the paper's primary result end to end. “an improved version of the gradient-boosting method LightGBM has been utilized to study the influencing factors of the four phase transformation tempe...”
Assessment of Support Vector Machine performance for default prediction and... Banks and Bank Systems 2022 11 assisted This is a human-run comparative assessment of an off-the-shelf technique's accuracy on an existing task (default prediction), where the SVM's predictions served to inform the authors' methodological comparison and tool design rather than constituting an AI-produced result that is itself the paper's contribution. “The results of this study showed that the model based on the Support Vector Machine approach with the Radial Basis Function kernel, performs better in...”
Application of Machine Learning Ensemble Methods to ASTRI Mini-Array Cherenk... Applied Sciences 2023 11 assisted The paper is a human-driven methodological comparison and tuning study of ML classifiers on simulated IACT data, where the models are evaluated components for the gamma/hadron separation pipeline rather than the AI itself producing the scientific result. “we have assessed and compared a number of supervised Machine Learning methods, including the Random Forest method, Extra Trees method, and Extreme Gra...”
Manifold Learning for Aerodynamic Shape Design Optimization Aerospace 2025 11 assisted The manifold-learning reduced-order model and autoencoder serve as fast surrogate components inside a human-developed genetic-algorithm optimization pipeline whose results are validated against full-order CFD, so the AI accelerates predictions within the workflow rather than itself performing the design optimization end to end. “the optimization outcomes aligned well with those obtained from the FOM-based framework, further validating the effectiveness and practical applicabil...”
Optimizing Crop Yield Prediction: Data-Driven Analysis and Machine Learning... Current Agriculture Research Journal 2024 10 assisted The paper's contribution is a human-run comparative evaluation of standard ML algorithms' accuracies on an existing crop-yield prediction task, not an AI system end-to-end solving the problem or producing the primary result itself. “We thoroughly assessed and contrasted the performance of a variety of machine learning algorithms, including Bagging Regressor, KNN, Decision Trees, G...”
Diagnosis of Pneumonia from Chest X-Ray Images Using Deep Learning venue not stated 2019 410 assisted The paper reports accuracy of two existing pre-trained CNNs on a chest X-ray dataset, explicitly framed as a computer-aided system to guide clinicians rather than an AI autonomously performing diagnosis. “computer-aided diagnosis systems are needed to guide the clinicians”
Identifying Exoplanets with Deep Learning: A Five-planet Resonant Chain arou... The Astronomical Journal 2018 368 assisted The neural network's role was to rank and classify candidate signals that the authors had already identified via their own search, and the authors themselves performed the statistical validation that confirmed the planets, making the AI a classification component in a human-run discovery pipeline rather than the end-to-end discoverer. “We statistically validate two new planets that are identified with high confidence by our model.”
Deep learning-based survival prediction of oral cancer patients Scientific Reports 2019 320 assisted The paper applies and validates an existing prediction algorithm's accuracy within a human-run retrospective clinical study, with the model's output intended to inform clinicians' treatment decisions rather than constituting an end-to-end AI-produced result. “deep learning-based survival prediction may improve prediction accuracy and guide clinicians both in choosing treatment options for better survival an...”
Multi-categorical deep learning neural network to classify retinal images: A... PLoS ONE 2017 282 assisted This is a pilot evaluation reporting a model's accuracy on an existing classification task (STARE fundus images), where performance was poor (30.5% accuracy for 10 classes) and the authors themselves conclude the techniques were ineffective, so the AI did not solve retinal disease detection but only produced evaluation results within a research workflow. “due to the small size of datasets, the deep learning techniques in this study were ineffective to be applied in clinics”
AutoGrow4: an open-source genetic algorithm for de novo drug design and lead... Journal of Cheminformatics 2020 171 assisted AutoGrow4 is a software tool for semi-automated, computer-AIDED drug discovery whose demonstrated output is only predicted candidate ligands (docking-scored, unvalidated) that feed into a human-run drug-development pipeline rather than an end-to-end solved drug design, and the paper's primary contribution is the toolkit itself. “We here present AutoGrow4, an open-source program for semi-automated computer-aided drug discovery.”
Machine learning explainability in nasopharyngeal cancer survival using LIME... Scientific Reports 2023 160 assisted The ML model is a prognostic decision-support tool whose survival predictions, validated on retrospective registry data and performing comparably to an existing XGBoost baseline, are framed as informing human treatment planning and clinical decisions rather than the AI itself producing a novel solved result. “ML may aid in planning individualized therapy for this patient population.”
Classification of Paediatric Inflammatory Bowel Disease using Machine Learni... Scientific Reports 2017 157 assisted The ML models serve as a diagnostic aid within a clinical workflow that physicians still complete, with the paper explicitly framing the modelling as supporting rather than performing diagnosis, and positioning itself as a 'blueprint' for ML use with clinical data. “This study employs mathematical modelling of endoscopic and histological data to aid diagnostic accuracy.”
Machine Learning of Designed Translational Control Allows Predictive Pathway... ACS Synthetic Biology 2018 118 assisted The ML models predict optimal RBS combinations from screened subsets, but the actual result, boosted production titers, was achieved through a human-run wet-lab pipeline of library screening, fermentation, and scale-up rescreening, making the ML a predictive component rather than the agent that produced the outcome. “our approach was able to boost production titers by over 60% when screening under 3% of a library”
Optimization of Silicone 3D Printing with Hierarchical Machine Learning 3D Printing and Additive Manufacturing 2019 111 assisted The HML model recommended optimal formulations and processing parameters within a human-run experimental pipeline (38 physical trial runs plus human printing and fidelity validation), so the AI informed an optimization workflow rather than performing the printing task itself. “the FRE method for 3D printing silicone parts was optimized based on a training set with 38 trial runs”
Maize yield forecasting by linear regression and artificial neural networks... The Journal of Agricultural Science 2014 100 assisted The ANN is one of several statistical tools (alongside MLR and persistence) inside a human-designed forecasting study whose contribution is a comparative skill evaluation and methodological findings, not an AI-produced result that itself constitutes the paper's achievement. “Forecasting the maize yield of China's Jilin province from 1962 to 2004... was investigated using multiple linear regression (MLR) and non-linear arti...”
RPITER: A Hierarchical Deep Learning Framework for ncRNA–Protein Interaction... International Journal of Molecular Sciences 2019 96 assisted The paper reports a prediction model's benchmark accuracy (AUC on five existing RPI datasets) as a computational aid for biological research, explicitly framing itself as a complementary tool rather than an agent that solved the biological problem of identifying ncRNA, protein interactions. “The proposed RPITER could be a complementary method for predicting RPI and constructing RPI network, which would help push forward the related biologi...”
Quantum machine learning for particle physics using a variational quantum cl... Journal of High Energy Physics 2021 87 assisted The paper's contribution is a human-designed hybrid variational quantum classifier benchmarked on a particle-physics classification task, where the classifier is one component inside a human-run resonance-search analysis rather than an AI that itself solved the physics problem. “By applying this algorithm to a resonance search in di-top final states, we find that this method has a better learning outcome than a classical neura...”
The Effects of Explanations in Automated Essay Scoring Systems on Student Tr... Journal of Learning Analytics 2023 76 assisted The paper's primary result is a human-factors finding about student trust and motivation, produced by a human-designed user study in which the automated essay scoring system was merely the stimulus/object under investigation rather than the agent producing the result. “The current study aimed to identify the effect of explanations of an automated essay scoring system on students' trust and motivation.”
Advancing Enzyme’s Stability and Catalytic Efficiency through Synergy of For... ACS Catalysis 2023 75 assisted The computational tools (FireProt, PROSS, machine learning) served as design-prediction components within a human-run protein engineering workflow that included manual curation, wet-lab construction, and crystallographic validation, rather than an AI itself producing the stabilized enzymes. “their prediction accuracy can be improved by applying manual curation or machine learning for the removal of potentially destabilizing mutations, yiel...”
Explaining deep neural networks for knowledge discovery in electrocardiogram... Scientific Reports 2021 75 assisted The paper presents an explainability method (ECGradCAM) whose attention maps are explicitly framed as aids for human clinicians and researchers who remain responsible for decisions and knowledge discovery, with the AI output informing humans rather than solving a task end-to-end. “Attention maps may be used in the clinic to aid diagnosis, discover new medical knowledge, and identify novel features and characteristics of medical...”
Laser-Induced Breakdown Spectroscopy Assisted by Machine Learning for Plasti... Atoms 2019 72 assisted The ML algorithms (PCA/LDA) function as data-analysis components inside a human-designed and human-run LIBS experimental workflow, reporting classification accuracies on curated lab samples as a technique demonstration, with the actual sorting/recycling application left prospective. “The combination of LIBS technique with these machine learning algorithmic approaches, in particular the latter, provided excellent classification resu...”
Machine-Learning-Guided Library Design Cycle for Directed Evolution of Enzym... ACS Catalysis 2021 70 assisted The ML only predicted variant libraries within a human-run directed evolution pipeline where humans performed the mutagenesis, experimental activity assays, and iterative wet-lab evaluation, so the improved enzymes were produced by the human workflow rather than by the AI end to end. “variants predicted by the first round were experimentally evaluated, and used as additional training data for the second-round prediction”