Deepscaler Easy: A Subset of Mathematical Reasoning Problems with High Pass Rate

A subset of the DeepScaleR dataset, specifically the 'easy' part where problems have a pass rate greater than 4 out of 8. The dataset was created by Taiqiang Wu and colleagues, with associated research published in 2026. It is hosted on Hugging Face and was last updated on 2026-02 25.

Use Cases

• Benchmarking model performance on easier mathematical reasoning problems based on the described pass rate filter.
• Training or fine-tuning language models for question answering tasks using curated text problems.
• Analyzing the difficulty spectrum within mathematical reasoning datasets based on the 'easy' subset selection.

Strengths

• Dataset is linked to a specific, cited research paper from 2026, providing academic context.
• Platform tags indicate a clear focus on question answering, mathematics, and reasoning tasks.
• Last update timestamp of 2026-02-25 is provided.

Limitations

• Row count, column definitions, and file formats are unknown, limiting suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Hugging Face user 'taki555', associated with the research paper 'The Art of Efficient Reasoning: Data, Reward, and Optimization'.

Collection Method

Likely curated or generated as part of the research project described in the citation.

Time Range

null

Freshness

Last updated 2026-02-25 08:53:31; freshness should be verified.

Geography

null