concept

ELI

Also known as: EAT-Lancet Index, ELI5

Facts (26)

Sources

How do the indices based on the EAT-Lancet recommendations ... medrxiv.org medRxiv May 14, 2024 20 facts

measurementThe EAT-Lancet Index (ELI), which uses semi-quantitative scores, shows positive associations with polyunsaturated fatty acids and vitamins D and E, but negative associations with protein, B-complex vitamins, phosphorus, calcium, and iron.

measurementWomen had significantly higher means in WISH, ELI, PHDI, and ELDS dietary index scores than men.

measurementThe WISH and ELI indices have moderate effects on the PANDiet index (η2 = 0.121 and 0.083, respectively), while the PHDI and ELD-I indices have small effects (η2 = 0.059 and 0.053, respectively).

measurementCorrelations between energy intake and other dietary indices were low (p < 0.0001): HSDI (r = -0.23), WISH (r = -0.25), ELI (r = -0.28), and ELDS (r = -0.30).

measurementThe ELD-I index had the highest split-half reliability (λ4 = 0.57), followed by WISH (λ4 = 0.47), ELI (λ4 = 0.47), and PHDI (λ4 = 0.46).

claimThe ELI index exhibited negative trends in GHGE, freshwater eutrophication, and freshwater ecotoxicity, with small effects on other environmental indicators.

measurementThe indices based on EAT-Lancet recommendations are generally negatively correlated with the Product Environmental Footprint (PEF) and 14 individual environmental metrics (highest ρ = −0.33), except for water use and photochemical ozone formation, which showed positive correlations for the WISH and ELI indices.

claimStructural Equation Modeling (SEM) confirmed the unidimensional structural validity of the WISH, PHDI, ELD-I, and ELI dietary indices.

measurementIndividuals in lower BMI groups had higher scores on dietary indices, though this trend was only confirmed for ELD-I and ELI.

claimThe ELI dietary index exhibits moderate-sized positive trends regarding DHA, EPA+DHA, and vitamin C, and negative trends regarding zinc.

measurementThe EAT-Lancet Index (ELI) assesses dietary intake in grams per day without energy adjustment, using a semi-quantitative scoring scale ranging from 0 (non-compliance) to 3 points (high compliance) per component, resulting in a total score range of 0 to 42 points.

measurementNon-smokers have higher mean scores for WISH, ELD-I, and ELI dietary indices.

measurementSignificant positive correlations were found between the PANDiet adequacy sub-score (Mean = 63.42; SD = 12.10) and the WISH, PHDI, ELD-I, and ELI indices, with ρ ranging from 0.07 to 0.17.

referenceThe EAT-Lancet Index (ELI) consists of 14 food components divided into 7 positive components (emphasized foods) and 7 negative components (limited foods).

measurementIndividuals with higher educational levels had higher mean scores in PHDI, ELD-I, and ELI dietary indices compared to those with lower formal education.

claimThe EAT-Lancet Diet Index (ELD-I) and the EAT-Lancet Index (ELI) show stronger correlations regarding convergent validity related to environmental impact.

measurementAmong the indices based on EAT-Lancet recommendations, the ELD-I and ELI indices show the highest correlations with environmental impact indicators, while the PHDI index shows the weakest (ρ < −0.10) and least significant coefficients.

measurementThe EAT-Lancet Index (ELI) mean score in the French study was similar to that of the Swedish study, although the specific food components contributing to the scores differed due to consumption pattern variations.

claimThe EAT-Lancet Index (ELI), which was developed using Swedish data, is associated with reduced mortality and a lower risk of chronic diseases.

measurementThe Swedish cohort used to design the EAT-Lancet Index (ELI) reported a vegetable consumption of less than 200 g/d, a potato consumption of more than 100 g/d, and a fish consumption of more than 50 g/d.

Real-Time Evaluation Models for RAG: Who Detects Hallucinations ... cleanlab.ai Cleanlab Apr 7, 2025 4 facts

claimIn the ELI5 benchmark, AI responses annotated as incorrect exhibit issues such as factual errors, misrepresentations, oversimplification that introduces inaccuracies, or claims not supported by the retrieved context.

claimIn the ELI5 benchmark, the Prometheus and TLM evaluation models are more effective at detecting incorrect AI responses than other detectors, though no method achieves very high precision or recall.

claimThe Cleanlab RAG benchmark reports results over 6 datasets, including four from the HaluBench suite and two datasets named FinQA and ELI5.

referenceThe ELI5 dataset captures the challenge of breaking down complicated concepts into simplified explanations without sacrificing accuracy.

EdinburghNLP/awesome-hallucination-detection - GitHub github.com GitHub 2 facts

referenceThe ASQA, QAMPARI, and ELI5 datasets are question-answering datasets characterized by containing factual questions where references are important, requiring long-text answers covering multiple aspects, and necessitating the synthesis of multiple sources.

referenceEvaluation metrics for AI systems include fluency (measured by MAUVE), correctness (measured by EM recall for ASQA, recall-5 for QAMPARI, and claim recall for ELI5), and citation quality (measured by citation recall and citation precision).