Fineweb Edu 100M: A 100 Million Token Sample of Educational Web Text

100 million tokens of educational web resources sampled from the HuggingFaceFW/fineweb-edu dataset using reservoir sampling. The dataset was created by codelion and was last updated on November 2, 2025. Reservoir sampling ensures each token from the source had an equal probability of inclusion, aiming for a representative sample.

Use Cases

• Training language models on educational content based on the described curated web resources.
• Conducting ablation studies on educational text corpora based on the sampling methodology.
• Analyzing linguistic patterns in educational web documents based on the described content.

Strengths

• 100 million tokens provide a substantial text corpus for analysis.
• Reservoir sampling methodology aims to create a statistically unbiased, representative sample of the source.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The full scale (row count) and file formats are unknown, which may limit suitability assessment.
• Last updated 2025-11-02 15:56:45; freshness should be verified.

Provenance

Source

HuggingFaceFW/fineweb-edu

Collection Method

Created using reservoir sampling from the source dataset.

Freshness

Last updated 2025-11-02 15:56:45.