An English subset of the data used to train the GTE family of multilingual text embedding models, as described in the mGTE technical report (Zhang et al., 2024). This dataset was reconstructed and curated by lightonai to reproduce the English portion of the original training recipe. The dataset page was last updated on April 21, 2026.
Use Cases
- Pre-training text embedding models based on the curated English data recipe.
- Benchmarking embedding model performance against the mGTE training methodology.
- Conducting ablation studies on the impact of specific English data sources for embedding quality.
Strengths
- Dataset is explicitly curated to follow the data recipe from a published technical report (Zhang et al., 2024).
- Focuses on the English-language component of a multilingual training corpus.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count and file size are unknown, which may limit suitability assessment.
- The description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- Reconstructed from the data sources described in the mGTE technical report.
- Collection Method
- Curated assembly of the English subset by lightonai.
- Time Range
- null
- Freshness
- Last updated 2026-04 21 13:52:13; freshness should be verified.
- Geography
- null