PUM-MATH: Pairwise Prefix Preference Examples for Math Problem Reasoning

PUM-MATH contains pairwise prefix preference examples for gain-based evaluation of LLM reasoning. The dataset is associated with the paper 'From Correctness to Utility: Gain-Based Prefix Evaluation for LLM Reasoning' and was authored by zhiqix. It was last updated on June 8, 2026.

Use Cases

• Evaluating LLM reasoning performance based on outcome-grounded prefix utility comparisons.
• Training or fine-tuning models to generate preferred reasoning prefixes for math problems.
• Benchmarking new evaluation metrics against the gain-based prefix utility framework.

Strengths

• Dataset is directly associated with a published research paper, providing academic context.
• Last update timestamp is precisely recorded as 2026-06-08 05:04:13.

Limitations

• Row count, file formats, and column-level documentation are unknown, limiting suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

zhiqix

Collection Method

Likely created for the associated research paper on gain-based prefix evaluation.

Freshness

Last updated 2026-06-08 05:04:13.