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
A Hybrid Flux Balance Analysis and Machine Learning Pipeline Elucidates Meta... iScience 2020 68 assisted The ML pipeline (FBA, PCA, k-means, LASSO) is one component in a human-run analysis workflow whose biological conclusions about cyanobacterial adaptation are drawn by the authors, not autonomously produced as the paper's primary result. “We then incorporate principal-component analysis, k-means clustering, and LASSO regularization to reduce dimensionality and extract key cross-omic fea...”
AptaNet as a deep learning approach for aptamer–protein interaction predicti... Scientific Reports 2021 65 assisted AptaNet is a predictive model evaluated only on a constructed benchmark (reporting accuracy metrics), with the abstract explicitly framing it as a tool that 'can help identify' novel pairs rather than an agent that discovered or validated any aptamer, protein interactions itself. “The results indicate that AptaNet can help identify novel aptamer–protein interacting pairs”
The major effects of health-related quality of life on 5-year survival predi... Scientific Reports 2020 63 assisted The ML models are a methodological component in a human-designed prognostic modeling study whose actual contribution is the analytical finding that adding HRQOL variables improves survival prediction accuracy, not an AI-produced result standing on its own. “The primary goal of this study was to evaluate the major roles of health-related quality of life (HRQOL) in a 5-year lung cancer survival prediction m...”
Sweet corn yield prediction using machine learning models and field-level da... Precision Agriculture 2023 60 assisted The ML models are tools inside a human-run analysis whose stated goals are to benchmark model performance and extract variable-importance insights from historical data, not to deliver an end-to-end solved prediction capability. “This study utilizes a historic US sweet corn dataset to: (a) evaluate machine learning model performances on sweet corn yield prediction and (b) ident...”
Macrocyclization of linear molecules by deep learning to facilitate macrocyc... Nature Communications 2023 56 assisted Macformer is a generative component whose outputs were then filtered by human-run docking and wet-lab validation to design JAK2 inhibitors, making the AI one step in a human-driven pipeline rather than the end-to-end solver. “We showcase the utility of Macformer when combined with molecular docking simulations and wet lab based experimental validation, by applying it to the...”
Automated detection of European wild mammal species in camera trap images wi... European Journal of Wildlife Research 2020 55 assisted The paper is a human-run usability test of an existing, pre-trained TensorFlow Hub model, reporting its accuracy figures and framing the outcome as 'potential' for speeding up human camera-trap analysis (with further training needed), rather than the AI itself delivering a solved capability. “we have shown the potential of an existing and pre-trained image classification model for wildlife animal detection, classification, and analysis”
Automated generation of discharge summaries: leveraging large language model... Scientific Reports 2025 53 assisted The LLM only produced editable draft summaries for physicians to review and correct within a human-run workflow (data manually extracted from EHRs, outputs rated and finalized by humans), and the paper itself frames its contribution as descriptive insights to guide future work rather than a demonstrated solved capability, with frequent errors and omissions reported. “the goal is to assist physicians with editable draft summaries”
Fabric Defect Detection in Real World Manufacturing Using Deep Learning Information 2024 53 assisted The paper's primary result is a performance comparison of off-the-shelf detection models (YOLOv8/v5, MobileNetV2-SSD) on a newly collected textile dataset, i.e., reported accuracy on an existing task where the model serves as a component to assist human quality inspection rather than the AI itself being the contribution that solved the manufacturing defect problem. “YOLOv8n achieved the highest performance, with a mAP of 84.8%, precision of 0.818, and recall of 0.839 across seven different defect classes.”
On Efficiency of Selected Machine Learning Algorithms for Intrusion Detectio... International Journal of Electronics and Telecommunications 2016 50 assisted The paper's contribution is a concept and evaluation environment in which humans test and compare ML algorithms' detection efficiency on generated SDN traffic, making the AI a component of a human-run benchmarking workflow rather than the agent that solved intrusion detection. “The article examines the efficiency of selected machine learning methods: Self Organizing Maps and Learning Vector Quantization and their enhanced ver...”
Pill Detection Model for Medicine Inspection Based on Deep Learning Chemosensors 2021 48 assisted The abstract reports a deep learning model's benchmark-style detection improvement over existing algorithms, with the AI explicitly framed as a component that will aid the human-run medicine inspection workflow rather than itself completing inspection end to end. “Our proposed method will help minimize problems, such as loss of productivity and human error, which occur while inspecting dispensed pills.”
Additive manufacturing: A machine learning model of process-structure-proper... Materials Science in Additive Manufacturing 2022 47 assisted The ML model is a predictive correlation component inside a human-driven experimental workflow (humans fabricated samples, characterized microstructure via SEM/EBSD/XRD, and ran Taguchi machining tests), and the reported result is the model's prediction accuracy on that human-collected data rather than the AI performing the task end to end. “It was found that the prediction accuracy of this new PSP linkage is extremely high (>99%, statistically significant at 95% confidence interval).”
Multi-Feature Classification of Multi-Sensor Satellite Imagery Based on Dual... Sensors 2018 46 assisted The random forest classifier is a standard off-the-shelf tool embedded in a human-designed feature-engineering and sensor-comparison workflow, and the paper's contribution is the comparative accuracy analysis of feature/sensor combinations rather than a new AI capability or an AI-produced result. “The classification was performed using a random forest (RF) method.”
A Machine Learning Classifier for Fast Radio Burst Detection at the VLBA Publications of the Astronomical Society of the Pacific 2016 44 assisted The classifier is a filtering component in a human-run candidate review pipeline, humans still review and decide on the most promising candidates, so the AI aids rather than completes the discovery workflow. “leaving only the 10%–20% most promising candidates to be reviewed by humans”
Automated visitor and wildlife monitoring with camera traps and machine lear... Remote Sensing in Ecology and Conservation 2023 42 assisted This is a validation study of an existing open-source model (MegaDetector) whose outputs are benchmarked against human manual classifications within a human-run camera-trap workflow, so the AI is an evaluated component of the pipeline rather than the agent that produced the paper's primary result. “The performance at detecting and counting humans, animals and vehicles is evaluated by comparing the detection results with manual classifications of...”
Effects of data transformation and model selection on feature importance in... Microbiome 2025 37 assisted The paper is a human-designed methodological assessment where existing ML models serve as evaluation tools across 24 datasets, and the contribution is guidance about transformations rather than any AI-produced result. “Our aim was to assess the impact of various data transformations on the accuracy, generalizability and feature selection by analysis using more than 8...”
Advanced UAV Design Optimization Through Deep Learning-Based Surrogate Model... Aerospace 2024 37 assisted The deep neural network serves as a surrogate model providing performance estimates inside a human-built design/optimization workflow (genetic algorithm, mission scenario selection, case study), rather than the AI itself producing the UAV design result end to end. “This work addresses these challenges by integrating deep neural networks with a multiobjective genetic algorithm to optimize UAV configurations.”
Machine learning for surrogate process models of bioproduction pathways Bioresource Technology 2022 36 assisted The ML surrogate models merely approximate outputs of human-built process simulations within a human-run technoeconomic/LCA workflow, and the paper is an evaluation of that potential rather than the AI producing a novel result end to end. “The results demonstrate that surrogate models can be an accurate and effective tool for approximating the cost, mass and energy balance outputs of mor...”
Why do users override alerts? Utilizing large language model to summarize co... Journal of the American Medical Informatics Association 2024 35 assisted GPT-4 produced summaries of alert comments, but the actual optimization of clinical decision support remains a human expert workflow that the AI output merely informs, and the paper's primary result is an expert-rated quality comparison rather than AI-driven CDS improvement. “These AI-generated summaries could provide CDS experts with a novel means of reviewing user comments to rapidly optimize CDS alerts both online and of...”
Machine learning phenomics (MLP) combining deep learning with time-lapse-mic... Scientific Reports 2022 33 assisted The deep learning component only extracts latent phenotypes from time-lapse microscopy within a human-run experimental pipeline (cell culture, gene knockdown, drug dosing, proteomics validation) whose biological findings are the actual contribution. “exploiting and combining the potential of time-lapse microscopy for cell behavior data acquisition and robust deep learning software architectures for...”
Machine Learning Techniques for Land Use/Land Cover Classification of Medium... Journal of Environmental Geography 2020 33 assisted The paper's contribution is a human-run comparative evaluation of off-the-shelf ML algorithms (random forest, SVM, ANN) applied to a satellite classification workflow, with the AI models serving as tools inside a human-designed and human-assessed pipeline rather than being the capability that constitutes the contribution. “Comparing the latest algorithms, our research aims to determine the best option to classify land use/land cover with special focus on temporary inunda...”
Deep Learning-Based PC Member Crack Detection and Quality Inspection Support... Applied Sciences 2022 32 assisted The AI performs crack detection as one component inside a human-run PC member quality inspection workflow, and the paper's reported contribution is a 'support' system yielding workflow time savings (13 min) rather than the AI solving the inspection task end to end. “Through the smart PC member quality inspection system proposed in this study, the time required for each phase of the existing PC member quality inspe...”
Development of postoperative delirium prediction models in patients undergoi... Scientific Reports 2023 30 assisted This is a human-run prospective clinical study in which standard ML algorithms serve as analysis tools to build a risk-stratification model whose predictions are meant to inform clinicians' decisions, with the paper reporting model accuracy (AUROC) on an existing prediction task rather than an AI-produced result solving the problem end to end. “We aimed to develop and validate machine learning models to predict post-cardiovascular surgery delirium.”
Being confident in confidence scores: calibration in deep learning models fo... Remote Sensing in Ecology and Conservation 2024 29 assisted The AI models are the object of study rather than the solver: humans ran the calibration analysis, applied temperature scaling, and drew the methodological conclusions, with the models' predictions serving as components in that human-driven evaluation. “Using a large and diverse European camera trap dataset, we investigate whether deep learning models for species classification in camera trap images a...”
Deep Learning Models for Poorly Differentiated Colorectal Adenocarcinoma Cla... Diagnostics 2021 28 assisted The models are explicitly positioned as diagnostic support tools whose classifications inform pathologists within a human-run clinical workflow, and the paper's contribution is a reported accuracy (AUC 0.94-0.98) on a task pathologists already perform rather than the AI autonomously completing the clinical diagnosis end to end. “it is an area that could benefit from AI-based tools to aid pathologists in their clinical workflows”
Enhancement of protein thermostability by three consecutive mutations using... Scientific Reports 2021 26 assisted The machine learning model only ranked candidate mutations in silico to guide which variants humans synthesized and assayed; the thermostability enhancement itself was achieved and validated through the human-run mutagenesis, screening, and wet-lab assay pipeline. “Among 7786 combinations ranked in silico, 20 promising candidates were selected and assayed.”
Machine learning for the discovery of molecular recognition based on single-... npj Computational Materials 2022 26 assisted The ML model analyzes a human-generated experimental library to understand feature correlations and predict sensor responses, reporting out-of-sample accuracy rather than performing molecular recognition end to end. “Out-of-sample analysis of our ML model showed significant correlations between model predictions and actual sensor responses for 6 out of 8 experiment...”
Learning cis-regulatory principles of ADAR-based RNA editing from CRISPR-med... Nature Communications 2021 25 assisted Machine learning (XGBoost) is one analytical component in a human-run CRISPR mutagenesis and deep-sequencing pipeline, modeling editing levels to inform biological understanding rather than performing the task end to end. “We use machine learning to integrate diverse RNA sequence and structure features to model editing levels measured by deep sequencing.”
Intelligent structural design of composite concrete‐encased steel columns ba... Structural Concrete 2025 25 assisted The paper's primary result is a benchmark-style accuracy comparison of a predictive model (BPNN tuned by DMP-PSO) on an existing task, axial capacity prediction from a human-curated database, with the AI serving as a prediction and optimization tool inside a human-run structural design workflow rather than end-to-end producing the constitutive result. “The predictive performance significantly outperforms that of BPNNs optimized using PSO, Bayesian optimization (BO), and grid search methods”
Comparing Performance of Machine Learning Algorithms for Default Risk Predic... TEM Journal 2021 24 assisted This is a human-run comparative evaluation of existing ML algorithms on an existing Lending Club dataset, reporting model accuracies rather than presenting an AI-produced solution or new capability, which per the rules makes it assisted. “The purpose of this research is to evaluate several popular machine learning algorithms for credit scoring for peer to peer lending.”
Learning and interpreting the gene regulatory grammar in a deep learning fra... PLoS Computational Biology 2020 23 assisted The ResNet learned simulated regulatory grammars on synthetic data designed by humans, and the paper's contribution is an interpretability framework and analysis of AI capabilities rather than the AI independently solving a real biological problem. “Our results provide a framework for interpreting the regulatory rules learned by ResNets, and they demonstrate that the ability and efficiency of ResN...”
Practical fluorescence reconstruction microscopy for large samples and low-m... PLoS Computational Biology 2020 23 assisted The paper's contribution is a human-run contextual evaluation of an existing AI technique (FRM), plus design guidelines and adoption code/manual, with the CNN as a component whose reconstructions the humans assessed rather than a novel result the AI itself produced. “we relate the conventional benchmarks and demonstrations to practical and familiar cell biology analyses to demonstrate that FRM should be judged in c...”
Survival Prediction in Gallbladder Cancer Using CT Based Machine Learning Frontiers in Oncology 2020 23 assisted The machine learning (LASSO/Random Forest) is one component in a human-run radiomics workflow, manual tumor segmentation, LIFEx feature extraction, and a final multivariate COX model combining clinical factors, reporting predictive accuracy (AUC 0.79/0.73) on an existing prognostication task rather than the AI itself producing the primary result. “After obtaining the pre-treatment CT images, manual segmentation of the tumor lesion was performed and LIFEx package was used to extract the tumor sig...”
Unsupervised Machine Learning for Exploratory Data Analysis of Exoplanet Tra... The Planetary Science Journal 2022 22 assisted This is a human-designed exploratory methodology demonstration on a synthetic benchmark dataset, where unsupervised ML techniques (PCA, K-means, manifold learning) serve as components whose outputs inform human interpretation and recommendations rather than the AI itself producing the paper's primary result. “To illustrate the proposed unsupervised methodology, we use a well-known public benchmark data set of synthetic transit spectra.”
Development and Optimization of VGF-GaAs Crystal Growth Process Using Data M... Crystals 2021 22 assisted The ML models analyzed CFD-generated simulation data to identify influential parameters and provide guidelines, but the actual crystal growth process optimization remains a human-driven downstream task that the AI output merely informs. “capable of providing guidelines for understanding the crystal growth process, which is a prerequisite for the growth of low-cost, high-quality bulk cr...”
Modeling Xanthan Gum Foam’s Material Properties Using Machine Learning Metho... Polymers 2024 21 assisted The primary result is the human-produced and human-tested xanthan gum foam material; the ML models only fit and predicted properties of the experimentally measured formulations, serving as an efficiency/analysis component in a human-run experimental workflow. “Five machine-learning-based methods ... were utilized to analyze the effects of the components used in the foam formulation. These models yielded accu...”
Automatic improvement of deep learning-based cell segmentation in time-lapse... Bioinformatics 2021 20 assisted The paper's contribution is a human-designed NAS methodology (search space and search strategy) whose resulting networks are benchmarked for accuracy on an existing challenge task, making the AI the method under evaluation in a human-run pipeline rather than the agent whose output constitutes the contribution. “Comprehensive evaluations on the 2D datasets from the cell tracking challenge demonstrate the competitiveness of the proposed method compared to the s...”
Prediction of Three-Directional Ground Reaction Forces during Walking Using... Sensors 2023 20 assisted The paper's primary contribution is a human-engineered shoe sole sensor system, with ML regression serving as a calibration/prediction step inside that human-run instrumentation workflow, validated against force plates, rather than an AI capability that itself constitutes the result. “A regression model to predict three-directional ground reaction forces (GRFs) from force sensor outputs was created using multiple linear regression a...”
Symbolic regression as a feature engineering method for machine and deep lea... Machine Learning Science and Technology 2024 19 assisted Symbolic regression is used as a feature-engineering component inside a human-run ML pipeline, with the paper's contribution being a method demonstration of performance gains rather than an AI system itself solving a problem end to end. “we propose to integrate symbolic regression (SR) as an FE process before a ML model to improve its performance”
Machine learning for combinatorial optimization of brace placement of steel... Japan Architectural Review 2018 19 assisted The ML component only pre-screens candidate solutions to skip costly structural analyses within a human-designed simulated annealing optimization pipeline, rather than itself producing the optimal brace placement. “the computational cost is successfully reduced by avoiding costly structural analysis for a solution judged by machine learning as nonoptimal”
Machine learning techniques for default prediction: an application to small... Risk Management 2023 19 assisted The ML models are evaluated as prediction components within a human-run comparative study of credit risk methods, with the paper's contribution being the performance comparison rather than the AI end-to-end solving default prediction. “This paper uses a large sample of small Italian companies to compare the performance of various machine learning classifiers and a more traditional lo...”