OpenDataArena-scored-data-2603: 63 Scored Instruction-Following Datasets

A collection of 63 high-quality instruction-following datasets containing nearly 25 million samples. Each sample is scored across 30 dimensions, including IFD, PPL, and Deita_Quality. The dataset was created by OpenDataArena and last updated on 2026-04-27.

Use Cases

• Filter training data based on the 30-dimensional scoring metrics mentioned in the description
• Implement curriculum learning strategies using the per-sample quality scores
• Optimize data mixtures for model training by leveraging the fine-grained sample evaluations

Strengths

• Contains nearly 25 million samples, providing a large-scale resource
• Each sample is evaluated on 30 distinct metrics, enabling detailed analysis
• Aggregates 63 high-quality instruction-following datasets into a single collection

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• License information is unavailable, which may restrict usage

Provenance

Source

OpenDataArena

Collection Method

Aggregated from multiple sources; scoring methodology not detailed in the provided description.

Time Range

null

Freshness

Last updated 2026-04-27 04:12:13; freshness should be verified

Geography

null