Try:droughtFusariumnitrogen fertilizerArabidopsisseed germination

“errors”

409 matches

Claim types

Direction of the findings

plus 189 non-evaluative

Over time

1950 ’60 ’70 ’80 ’90 2000 ’10 ’20

plus 6 earlier · 1920s–1930s

Refine ·

More filters

Clear filters

Method / resourceobservedoriginal resultmeasured

Claim · model-written summary

For many whitefly detection models, most errors were species classification errors.

Taxon: tomatoSolanum lycopersicumt0
Outcome: most errors occurred in species classificatione0
Comparator: many modelse0

Verbatim

e0
“While, for many models, most errors occurred in species classification, those trained on low-resolution images had a higher proportion of localization errors.”
✓ located verbatim in the source · abstract, characters 853–1011

t0
“tomato leaflets”
✓ located verbatim in the source · title, characters 98–113

Title

Effect of image resolution on automatic detection of whitefly (Hemiptera: Aleyrodidae) species on tomato leaflets using deep learning

Mikio Kamei Smart Agricultural Technology 2023. doi:10.1016/j.atech.2023.100372

Other claims from this paper (7)

Method / resource Whitefly detection performed well with high-resolution training and testing images.
Method / resource Whitefly detection performance declined as image resolution decreased.
Method / resource Low-resolution whitefly models had a higher proportion of localization errors.
Method / resource Mismatch in resolution between training and inference produced poor whitefly detection performance.
Method / resource Mixed-resolution training gave similar whitefly detection performance to single-resolution training.
Method / resource detectoRS performed best for lower- or mixed-resolution whitefly images.
Method / resource YOLOX had the highest whitefly detection accuracy for high-resolution images.

Method / resourceobservedoriginal resultpositivemeasured

Claim · model-written summary

Semantic anchoring constrained predictions under related errors.

Outcome: predictions under related errorse0
Comparator: semantic anchoringe0

Not reportedTaxon

Verbatim

e0
“semantic anchoring constrains predictions under related errors”
✓ located verbatim in the source · abstract, characters 890–952

Title

SMART: Structured multimodal agricultural retrieval-augmented transformer for plant disease diagnosis

Yuan-Chia Chan; Yao-Chung Fan; Hung-Chung Li; et al. Computers and Electronics in Agriculture 2026. doi:10.1016/j.compag.2026.111882

Other claims from this paper (8)

Method / resource Structured captions outperformed LLM-generated narratives for disease diagnosis.
Method / resource The vision-only baseline exceeded all LLM approaches on the 20-class pilot.
Method / resource Asymmetric loss design mitigated long-tailed degradation.
Method / resource The retrieval-based safety net bounded degradation under noisy inputs.
Method / resource Semantic anchoring benefited high-frequency classes.
Method / resource The retrieval-based safety net preferentially compensated rare classes.
Method / resource SMART achieved F1 of 0.9767 versus the best fine-tuned CNN F1 of 0.9118.
Method / resource SMART achieved F1 of 0.9767 versus the best multimodal foundation model F1 of 0.8904.

Observationobservedoriginal resultmeasured

Claim · model-written summary

Farmers made 22 errors in red palm weevil control of date palms, grouped into delayed-control, horticultural and treatment errors.

Taxon: date palmsPhoenix dactyliferae6
Outcome: errorse1
Value: 22e1

Verbatim

e1
“A total of 22 errors were identified and categorized into three groups”
✓ located verbatim in the source · abstract, characters 257–327

e6
“a sector with 32,000 date palms (El Kharga 10) was selected”
✓ located verbatim in the source · abstract, characters 1524–1583

Title

The Positive Impact of Correcting Date Palm Farmers' Mistakes That Hinder the Control of the Red Palm Weevil (Rhynchophorus ferrugineus, Olivier)

Gameel S. M.M.; Amany O. M. Osman; Aya S. Gameel Egyptian International Journal of Palms 2025. doi:10.21608/esjp.2025.409387

Other claims from this paper (8)

Intervention effect The red palm weevil infestation rate in 2021 was 0.7%, down from 1% in 2020, after farmers were trained.
Intervention effect The red palm weevil infestation rate in 2022 was 0.47%, down from 1% in 2020, after farmers were trained.
Intervention effect The red palm weevil infestation rate in 2023 was 0.16%, down from 1% in 2020, after farmers were trained.
Intervention effect Infestation reduction in 2021 was 0.30% after farmers were trained.
Intervention effect Infestation reduction in 2022 was 0.53% after farmers were trained.
Intervention effect Infestation reduction in 2023 was 0.84% after farmers were trained.
Observation The primary reasons for delayed control were farmers lacking awareness of pest symptoms and stages, relying on the Agriculture Directorate and having limited financial resources.
Association Incorrect pesticide application, excessive concentrations and improper injections at the terminal bud resulted in palm mortality in date palms.

Method / resourceobservedoriginal resultmeasured

Claim · model-written summary

In second validation, 54% of scenarios had errors within 10%.

Taxon: SOYBEANGlycine maxt0
Outcome: errors within 10%e0
Value: 54%e0

Verbatim

e0
“54% of the validation scenarios had errors within 10%”
✓ located verbatim in the source · abstract, characters 2127–2180

t0
“SOYBEAN RUST”
✓ located verbatim in the source · title, characters 36–48

Title

DEVELOPMENT OF A NEURAL NETWORK FOR SOYBEAN RUST EPIDEMICS

W. D. Batchelor; X. B. Yang; A. T. Tschanz Transactions of the ASAE 1997. doi:10.13031/2013.21237

Other claims from this paper (9)

Method / resource With independent epidemics, the best four-input, four-hidden-node network achieved R2 0.793 for validation scenarios.
Method / resource The same configuration had an average error of 11.0% per observation.
Method / resource Under independent-epidemic validation, 50% of scenarios had prediction errors within 10% of measured rust severity.
Method / resource Under independent-epidemic validation, 80% of scenarios had prediction errors within 20% of measured rust severity.
Method / resource With 1980 epidemics for training and 1981 epidemics for validation, three- and four-input networks gave very similar results.
Method / resource The three-input network achieved R2 0.858 in the second validation.
Method / resource The three-input network had an average error of 9.7% in the second validation.
Method / resource The three-input network had a maximum validation error of 43%.
Method / resource In second validation, 84% of scenarios had errors within 20% of measured values.

Associationobservedoriginal resultmeasured

Claim · model-written summary

Mass loss measurement errors were greatest at lower mass losses.

Taxon: birche1
Outcome: errorse0
Direction: greateste0
Comparator: lower mass lossese0

Verbatim

e0
“Such errors are greatest at lower mass losses.”
✓ located verbatim in the source · abstract, characters 648–694

e1
“birch blocks contained 31.3% fungal biomass”
✓ located verbatim in the source · abstract, characters 344–387

Title

Fungal biomass in decayed wood

Heather L. Jones; James J. Worrall Mycologia 1995. doi:10.1080/00275514.1995.12026555

Other claims from this paper (4)

Observation Birch blocks decayed by T. versicolor contained 31.3% fungal biomass after 12 weeks, the highest fungus wood combination.
Observation Fungal biomass in birch decayed by B. adusta exceeded 9.0%.
Observation Fungal biomass in birch decayed by G. trabeum exceeded 9.0%.
Method / resource Apparent mass loss underestimated corrected wood mass loss by as much as 42%.

Method / resourceobservedoriginal resultpositivemeasured

Claim · model-written summary

Thermal image classification had fewer errors than normal image classification.

Outcome: errorse0
Direction: moree0
Comparator: Normal image classificatione0

Not reportedTaxon

Verbatim

e0
“Normal image classification causes more errors than thermal image classification”
✓ located verbatim in the source · abstract, characters 483–563

Title

ExplainableAI with Transfer Learning for Crop Disease Assessment Using Thermal Imaging

Suryanarayanan V; Praveen Sankar; Shankaranarayanan V; et al. 2025 International Conference on Data Science, Agents & Artificial Intelligence (ICDSAAI) 2025. doi:10.1109/icdsaai65575.2025.11011583

Other claims from this paper (1)

Method / resource EfficientNetB0 with XAI gave higher accuracy than traditional models.

Method / resourceobservedoriginal resultpositivemeasured

Claim · model-written summary

Confusion matrix results showed high accuracy with few classification errors.

Taxon: SoybeanGlycine maxt0
Outcome: accuracy with few errorse0

Not reportedComparatorValue

Verbatim

e0
“The confusion matrix showed high accuracy with few errors.”
✓ located verbatim in the source · abstract, characters 1011–1069

t0
“Soybean”
✓ located verbatim in the source · title, characters 51–58

Title

Fast and Efficient CNN Architectures for Automated Soybean Leaf Disease Classification

Kalpesh Patel; Atul Patel International Journal of Basic and Applied Sciences 2025. doi:10.14419/myqfsc31

Other claims from this paper (3)

Method / resource Enhanced CNN attained 98.14% test accuracy and outperformed three reference models.
Method / resource Fast CNN attained 95.70% accuracy with reduced training time.
Recommendation The models may support rapid disease management and agricultural AI tools.

Method / resourceobservedoriginal resultmeasured

Claim · model-written summary

Under independent-epidemic validation, 80% of scenarios had prediction errors within 20% of measured rust severity.

Taxon: SOYBEANGlycine maxt0
Outcome: errors within 20%e0
Value: 80%e0

Verbatim

e0
“80% had errors within 20%”
✓ located verbatim in the source · abstract, characters 1491–1516

t0
“SOYBEAN RUST”
✓ located verbatim in the source · title, characters 36–48

Title

DEVELOPMENT OF A NEURAL NETWORK FOR SOYBEAN RUST EPIDEMICS

W. D. Batchelor; X. B. Yang; A. T. Tschanz Transactions of the ASAE 1997. doi:10.13031/2013.21237

Other claims from this paper (9)

Method / resource With independent epidemics, the best four-input, four-hidden-node network achieved R2 0.793 for validation scenarios.
Method / resource The same configuration had an average error of 11.0% per observation.
Method / resource Under independent-epidemic validation, 50% of scenarios had prediction errors within 10% of measured rust severity.
Method / resource With 1980 epidemics for training and 1981 epidemics for validation, three- and four-input networks gave very similar results.
Method / resource The three-input network achieved R2 0.858 in the second validation.
Method / resource The three-input network had an average error of 9.7% in the second validation.
Method / resource The three-input network had a maximum validation error of 43%.
Method / resource In second validation, 54% of scenarios had errors within 10%.
Method / resource In second validation, 84% of scenarios had errors within 20% of measured values.

Method / resourceobservedoriginal resultpositivemeasured

Claim · model-written summary

VGG16 made only 12 misclassification errors out of 357 samples in distinguishing healthy and infected plants.

Outcome: errorse0
Value: 12 errors out of 357 samplese0

Verbatim

e0
“only 12 errors out of 357 samples”
✓ located verbatim in the source · abstract, characters 854–887

Title

Integrating deep learning and field validation into a decision support system for Northern Corn Leaf Blight management in maize

Jadesha G.; Anurag Dhole; Deepak D.; et al. BMC Plant Biology 2026. doi:10.1186/s12870-026-08967-z

Other claims from this paper (8)

Method / resource VGG16 outperformed all other models evaluated for NCLB classification, achieving 94.0% accuracy.
Method / resource VGG16 had balanced precision, recall and F1-score of 0.94.
Method / resource VGG16 had an AUC-ROC of 0.93.
Intervention effect DSS-guided fungicide recommendations reduced disease incidence in maize to 6.8%, compared with 67.4% in controls.
Intervention effect DSS-guided fungicide recommendations achieved 90% disease reduction in maize.
Intervention effect DSS-guided fungicide recommendations increased maize grain yield by 35.4% (8.55 t/ha).
Observation The DSS-guided management had a favorable cost-benefit ratio of 1:2.49.
Intervention effect Cumulative disease burden was reduced by approximately 85% with DSS compared with the untreated control.

Associationobservedoriginal resultmeasured

Claim · model-written summary

As patch size increased, GPS area estimation errors decreased.

Taxon: hemp dogbaneApocynum cannabinume2
Treatment: patch sizee3
Outcome: errorse3
Direction: decreasede3

Verbatim

e2
“applied to natural infestations of hemp dogbane”
✓ located verbatim in the source · abstract, characters 1028–1075

e3
“As patch size increased, errors decreased”
✓ located verbatim in the source · abstract, characters 499–540

Title

Accuracy of a Global Positioning System (GPS) for Weed Mapping

Theodore M. Webster; John Cardina Weed Technology 1997. doi:10.1017/s0890037x00043438

Other claims from this paper (8)

Method / resource GPS area estimates for 5 m2 patches had errors of 7 to 45 per cent relative to actual measurements.
Method / resource GPS area estimates for 50 m2 patches had errors of 6 to 15 per cent relative to actual measurements.
Method / resource GPS area estimates for 500 m2 patches had errors of 3 to 6 per cent relative to actual measurements.
Association A curve relating GPS error to patch size had an excellent fit (r2 = 0.92).
Association GPS measurement error remained the same in absolute terms across all patch sizes.
Method / resource The GPS had submeter accuracy in navigation to the correct quadrat 73 per cent of the time.
Method / resource The GPS located the correct quadrat 27 per cent of the time.
Method / resource The GPS invariably navigated to within 1.58 m of the correct quadrat.

Lexical search (BM25). All words must match; if nothing does, any word. Organism names also match every spelling of the same organism. No semantic search yet.

“errors”

Claim types

Direction of the findings

Over time

Claim type

Research area

Polarity

Statistics