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
3Cnet: pathogenicity prediction of human variants using multitask learning w... Bioinformatics 2021 39 assisted 3Cnet is a pathogenicity-prediction tool whose predictions serve as input to a human-run clinical diagnostic workflow (the paper frames its value as improving diagnosis rates), rather than the AI itself producing the diagnostic result. “more effectively discovering pathogenic variants and thereby improving diagnosis rates”
Deep learning for dual detection of microsatellite instability and POLE muta... npj Precision Oncology 2024 38 assisted The AI's image-based predictions are explicitly positioned as a preliminary screening step whose flags inform a clinical workflow that humans still complete via confirmatory molecular testing, and the POLE result is a 75% screening flag rather than an end-to-end detection. “deploying this DL model as a preliminary screening tool could facilitate the efficient identification of clinically relevant MSI and POLE mutations in...”
Machine Learning Explainability in Breast Cancer Survival Studies in health technology and informatics 2020 38 assisted The ML survival model is a component in a human-driven workflow whose actual contribution is the authors' comparison of LIME and SHAP explanations, not an AI-produced result standing as the paper's primary outcome. “we used Local Interpretable Model-Agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP) to interpret the model's predictions”
Deep learning for prediction of energy consumption: an applied use case in a... Applied Intelligence 2024 37 assisted The paper's contribution is a comparative accuracy evaluation of existing neural architectures on a prediction task that is explicitly framed as one step inside a human-driven energy-management workflow, not an AI end-to-end solving of the problem. “performed a comprehensive evaluation of the accuracy of feed-forward and recurrent neural networks to predict energy consumption”
Deep Learning-Based Modeling of Drug–Target Interaction Prediction Incorpora... Interdisciplinary Sciences Computational Life Sciences 2023 36 assisted The abstract only reports a new model's comparative benchmark accuracy (MSE/AUPR) on the existing drug-target interaction prediction task, using expert-derived binding-site features, with predictions meant to inform human drug-discovery workflows rather than the AI itself producing the solved result. “The proposed method achieved better or similar performance on the MSE and AUPR metrics than the shallow methods.”
A Goldilocks Principle for the Gut Microbiome: Taxonomic Resolution Matters... mBio 2022 35 assisted The machine learning models serve as measurement instruments within a human-designed comparative study; the paper's contribution is the meta-finding about optimal taxonomic resolution, not an AI-produced detection capability, making this a benchmark-style analysis where AI informed the human researchers' conclusion. “To address this question, we used a reproducible machine learning framework to quantify classification performance of models based on data annotated t...”
A machine learning approach for multimodal data fusion for survival predicti... npj Precision Oncology 2025 35 assisted The paper's contribution is a reusable comparison pipeline and benchmark-style accuracy findings on an existing task (survival prediction on public TCGA data), with ML models as evaluated components rather than an AI-produced solution to the problem itself. “The study provides a reusable pipeline for the research community, suggesting future work on larger cohorts.”
A novel approach for skin lesion symmetry classification with a deep learnin... Computers in Biology and Medicine 2022 35 assisted The paper's contribution is a human-designed and human-evaluated CNN classification pipeline whose symmetry predictions are meant to inform dermatologists' lesion assessment, with the result being a benchmark F1-score on a human-labeled dataset rather than an AI-produced outcome. “we present a new simple, robust and fast classification pipeline that outperforms methods based on traditional approaches or pre-trained networks, wit...”
Classifying and diagnosing Alzheimer’s disease with deep learning using 6735... Scientific Reports 2025 33 assisted The paper reports accuracy metrics of four off-the-shelf pretrained CNN architectures fine-tuned on an existing public MRI dataset, which is benchmark-style performance reporting on an existing task rather than the AI itself solving Alzheimer's diagnosis end to end. “In this study, the Alzheimer MRI Preprocessed Dataset was used, which includes 6735 brain structural MRI scan images. After data preprocessing and nor...”
Interpretable machine learning for prediction of clinical outcomes in acute... Frontiers in Neurology 2023 33 assisted This is a routine clinical prediction model paper: the ML system reports accuracy (AUC) on an existing prognostication task using retrospective registry data, and its output is designed to inform human clinicians who make the actual treatment decisions, so the AI is a decision-support component rather than the agent that solved a problem. “Predicting the prognosis of acute ischemic stroke (AIS) is crucial in a clinical setting for establishing suitable treatment plans.”
Ensemble-Based Machine Learning Algorithm for Loan Default Risk Prediction Mathematics 2024 31 assisted This is a human-run comparative benchmark study applying off-the-shelf ML models and resampling techniques to an existing task (LendingClub default prediction), reporting accuracy metrics rather than an AI system that itself solved a problem end to end. “Our proposed model achieved an accuracy of 93.7%, a precision of 95.6% and a recall of 95.5%”
Credit risk prediction based on causal machine learning: Bayesian network le... Journal of Forecasting 2024 30 assisted The paper reports a Bayesian network model's predictive accuracy on six existing labeled credit datasets as a comparative method evaluation, and explicitly frames its output as decision support informing human risk judgments rather than as an AI-produced primary result. “This analysis provides decision‐makers with the necessary tools to make informed judgments about the risk profile of borrowers.”
Dynamic Learning Rate in Deep CNN Model for Metastasis Detection and Classif... Computational and Mathematical Methods in Medicine 2021 30 assisted The abstract explicitly frames the model as a second-opinion CAD tool assisting medical experts, and the paper's contribution is a training hyperparameter (dynamic learning rate) demonstrated via accuracy on an existing classification task, not an AI end-to-end solving of metastasis detection. “Computer-aided diagnosis (CAD) assists medical experts as a second opinion tool in early detection to prevent further proliferation.”
Prediction and risk assessment of sepsis-associated encephalopathy in ICU ba... Scientific Reports 2022 30 assisted The ML model is a clinical prediction tool whose output informs physicians' SAE risk assessment rather than performing the diagnostic task end-to-end, with physicians explicitly evaluating and using the model's predictions. “The interpretable model was effectively evaluated by professional physicians and can help them predict the occurrence of SAE more intuitively.”
A Deep Learning Model for Prostate Adenocarcinoma Classification in Needle B... Diagnostics 2022 30 assisted The model reports classification accuracy (ROC-AUC) as a screening aid explicitly framed to assist pathologists in the diagnostic workflow, rather than producing a standalone diagnostic result end-to-end. “Computational pathology applications that can assist pathologists in detecting and classifying prostate adenocarcinoma from whole-slide images (WSIs)...”
Evaluating Diagnostic Accuracy and Treatment Efficacy in Mental Health: A Co... European Journal of Investigation in Health Psychology and Education 2025 29 assisted This is an evaluation study benchmarking existing commercial LLMs' diagnostic accuracy against professional norms, and its own conclusion frames the models as supportive tools requiring human oversight rather than systems that solved the clinical task. “LLMs provide valuable support in diagnostics and treatment planning but cannot replace professional discretion.”
Weed Detection from Unmanned Aerial Vehicle Imagery Using Deep Learning—A Co... Sensors 2024 29 assisted The paper's contribution is a low-cost multispectral camera hardware comparison, with an off-the-shelf U-Net serving as the evaluation instrument inside a human-run pipeline (sensor building, data collection, thresholding/annotation, training, evaluation), so the AI informed rather than itself constituted the solved problem. “This research paper introduces a self-built, low-cost, multispectral camera system and evaluates it against the high-end MicaSense Altum system.”
A Hybrid Semi-Automated Workflow for Systematic and Literature Review Proces... Future Internet 2024 28 assisted The LLM only summarizes and extracts information that a researcher then uses to make inclusion/exclusion decisions in an explicitly semi-automated, human-in-the-loop workflow, so the AI is a component in a human-run process rather than the agent producing the review outcome. “This is then used by a researcher to make inclusion/exclusion decisions quickly.”
Interpretable machine learning for materials discovery: Predicting CO2 adsor... APL Materials 2024 28 assisted The ML model improves prediction accuracy and interpretability of CO2 adsorption on an existing hypothetical MOF dataset, serving as a screening guide within a human-run discovery pipeline rather than itself discovering or validating a new material. “Overall, this study provides a novel perspective to guide the initial high-throughput screening of MOFs as high-performance CO2 adsorption materials.”
Validating the accuracy of deep learning for the diagnosis of pneumonia on c... European Radiology Experimental 2024 25 assisted This is a human-run validation study in which a pre-existing CNN's CXR predictions were benchmarked against a reference diagnosis produced by a multidisciplinary expert panel; the paper's contribution is the evaluation finding (that AI accuracy was previously overestimated), not an AI-produced solution to the diagnostic problem. “we evaluated the accuracy of DL in diagnosing pneumonia on CXR using a more robust reference diagnosis”
A Deep CNN Technique for Detection of Breast Cancer Using Histopathology Ima... Advanced Computing and Communication Technologies for High Performance Applications (ACCTHPA) 2020 24 assisted The CNN is presented as a diagnostic aid to 'reduce the burden from doctors,' reporting benchmark classification accuracy on an existing dataset (BreakHis) rather than the AI itself producing an end-to-end cancer detection result. “To enhance efficiency of BC i.e. Breast Cancer detection using histopathology images and also to reduce the burden from doctors, we design a deep lear...”
A Hybrid Deep Reinforcement Learning and Optimal Control Architecture for Au... Energies 2023 23 assisted The DRL is only the high-level tactical layer of a human-designed hybrid architecture whose low-level control is handled by model-based NMPC, and the whole system is validated only in simulation rather than the AI itself being the end-to-end driving capability that constitutes the paper's result. “the DRL-based high-level planner is responsible for training the vehicle to choose tactical behaviors according to the surrounding environment, while...”
Monitoring hydropower reliability in Malawi with satellite data and machine... Environmental Research Letters 2020 23 assisted The random forest regression is one component inside a human-designed framework; its discharge predictions feed into a human-run assessment of hydropower reliability rather than the AI itself producing the paper's primary result. “The validated predictions are used to assess the impact of hydroclimatic extremes on hydropower reliability and on the final use of electricity in urb...”
Deep learning enhanced light sheet fluorescence microscopy for in vivo 4D im... Light Science & Applications 2025 22 assisted The AI (UI-Trans) is a restoration component inside a human-run light sheet microscopy pipeline, the microscope hardware acquires the low-exposure 4D data and the network denoises it, so the imaging result comes from the combined system rather than the AI alone. “we present a deep-learning enhanced light sheet fluorescence microscopy (LSFM) approach that addresses the restoration of rapid volumetric time-lapse...”
Implementation of large language models in electronic health records PLOS Digital Health 2025 22 assisted The paper reports a deployment and adoption study of an LLM assistant embedded in clinicians' workflows (summarization, retrieval, note drafting) where the AI's output supports humans who still perform the clinical work, rather than the AI itself producing the primary result. “The most common tasks were summarization, information retrieval, and note drafting, which together accounted for over 70% of interactions.”
Learning to quantify uncertainty in off-target activity for CRISPR guide RNA... Nucleic Acids Research 2024 21 assisted crispAI is an in-silico predictor whose uncertainty estimates and efficiency scores serve as decision support for human sgRNA designers, with the paper's results being benchmark accuracy/calibration comparisons against existing methods rather than the AI itself producing a completed outcome. “providing a more comprehensive risk assessment and facilitating improved decision-making in single guide RNA (sgRNA) design”
A Customized Multi‐Scale Deep Learning Framework for Storm Nowcasting Geophysical Research Letters 2023 21 assisted The paper's contribution is a plug-in methodological framework whose value is demonstrated through improved skill scores on the existing SEVIR nowcasting benchmark, a methods-improvement/benchmark-accuracy paper rather than the AI system itself delivering nowcasting as the paper's primary result. “The quantitative evaluation shows that all models applying our framework can improve skill scores by 8.5%–42.6% for 1‐hr nowcasting.”
Deep Learning for Line Intensity Mapping Observations: Information Extractio... The Astrophysical Journal Letters 2021 18 assisted The cGAN demonstrates a denoising/reconstruction capability entirely on simulated mock observation maps rather than solving the actual LIM observational problem, making it a method-development study whose AI output would inform future real-data analysis. “We train the cGANs using 30,000 mock observation maps with assuming a Gaussian noise matched to the expected noise level of NASA's SPHEREx mission.”
Machine learning enabled discovery of application dependent design principle... Machine Learning Science and Technology 2020 18 assisted The neural networks served as a property-prediction surrogate enabling a human-run screening pipeline, and the paper's actual contribution, the design principles, was derived by the authors through statistical analysis of the AI-screened candidates rather than produced by the AI itself. “By collecting statistics of the screened candidates, we investigate structural and compositional design principles that impact the properties of the s...”
Do Machine Learning Approaches Offer Skill Improvement for Short-Term Foreca... Weather and Forecasting 2022 16 assisted This is a human-designed systematic skill assessment of ANN models on ERA5 reanalysis data, where the AI is the object of evaluation rather than the agent producing the paper's result, and the models explicitly fail to fully solve the task (50% underestimation of strong gusts). “This study systematically assesses the skill enhancement that can be achieved using artificial neural networks (ANNs) for forecasting of wind gust occ...”
Leveraging Large Language Models for Enhancing Literature-Based Discovery Big Data and Cognitive Computing 2024 15 assisted The paper's contribution is a human-designed LBD framework/methodology where LLMs serve as enhancement components producing untested candidate hypotheses (on known illustrative scenarios like garlic and blood pressure) that would inform downstream human research, not an AI-produced discovery result. “Empirical validations, including scenarios on the effects of garlic on blood pressure and nutritional supplements on health outcomes, demonstrate the...”
Temporal and spatial satellite data augmentation for deep learning-based rai... Journal of Hydroinformatics 2024 15 assisted The paper's contribution is a human-designed comparative assessment of data augmentation techniques on an existing ConvLSTM nowcasting pipeline, with the AI model serving as the prediction instrument rather than being the demonstrated capability itself. “The goal of the study is to assess the impact of data augmentation on flood nowcasting.”
Integrating Satellite Imagery and Ground-Based Measurements with a Machine L... Hydrology 2023 15 assisted The random forest classifier is one component in a human-designed remote sensing workflow, humans selected the sites, assembled the satellite/in-situ data, and performed the trend and hydro-climatological factor analysis that constitutes the paper's actual findings, so the AI-assisted delineation only informed a human-completed monitoring study. “The random forest (RF) supervised classification algorithm was applied to estimate the lake area extent using Landsat imagery that was acquired betwee...”
GSpyNetTree: a signal-vs-glitch classifier for gravitational-wave event cand... Classical and Quantum Gravity 2024 14 assisted The paper presents a CNN classifier tool designed as one automation component within the human-run LIGO-Virgo candidate validation pipeline, rather than an AI-produced result where the classification itself is the solved problem. “LIGO-Virgo GW event candidate validation will require increased levels of automation”
An Automatic Source Code Vulnerability Detection Approach Based on KELM Security and Communication Networks 2021 12 assisted The abstract reports a model's comparative performance metrics (training speed, precision vs. Bi-LSTM/word2vec) on the existing task of vulnerability detection, making the AI a detection component in a workflow rather than an agent that itself produced a novel detection result. “Experimental results show that doc2vec vector representation brings faster training and better generalizing performance than word2vec. ELM converges m...”
SciDaSynth: Interactive Structured Data Extraction From Scientific Literatur... Campbell Systematic Reviews 2025 12 assisted SciDaSynth is an interactive human, AI collaborative tool in which the LLM generates candidate tables but researchers query, validate, and refine the outputs, so the structured data was produced through a human-run workflow rather than by the AI end-to-end. “A within‐subjects study with nutrition and NLP researchers demonstrates SciDaSynth's effectiveness in producing high‐quality structured data more effi...”
Machine Learning‐Driven Discovery of Thermoset Shape Memory Polymers With Hi... Journal of Polymer Science 2024 10 assisted The AI (VAE plus SVM) served as a predictive/screening component in a human-run design workflow, humans performed systematic molar ratio adjustments and used the model's fingerprints and Tg predictions to select candidates, rather than the AI itself end-to-end producing a validated discovery. “Through systematic molar ratio adjustments and VAE‐based fingerprinting, we discover novel SMP candidates with Tg values between 190°C and 200°C”
Discovery of EP4 antagonists with image-guided explainable deep learning wor... National Science Open 2025 1 assisted The deep learning workflow prioritized candidate EP4 antagonists, but the actual discovery, synthesis and wet-lab confirmation of nanomolar activity, selectivity, and macrophage assays, was completed by humans, making the AI a component in a human-run drug discovery pipeline. “Using LeadDisFlow, we identified four potent antagonists with single-nanomolar antagonistic activity against PGE2 receptor subtype 4 (EP4)”
Explainable machine learning can outperform Cox regression predictions and p... Scientific Reports 2021 335 assisted This is a human-run comparative evaluation that reports c-index accuracy of off-the-shelf ML models against Cox regression on an existing prediction task; the AI's outputs informed the authors' methodological conclusions rather than the AI itself producing the paper's result as a new end-to-end capability. “We demonstrated that in our dataset, ML-based models can perform at least as good as the classical CPH regression”
Multimodal skin lesion classification using deep learning Experimental Dermatology 2018 323 assisted The paper's contribution is a method (multimodal fusion) that improves a classifier's benchmark accuracy on an existing diagnostic task, positioning the AI as a clinical decision-support component whose performance is evaluated against human-provided labels rather than an AI system itself delivering the solved result. “We showed that our multimodal classifier outperforms a baseline classifier that only uses a single macroscopic image in both binary melanoma detection...”