Sign in to view source links and access this dataset
Description
520,000 error traces document models' mistakes during math problem synthesis. The dataset includes updates from March 2026, where 50,000 new datapoints were added and 10,000 older ones were replaced with higher-quality synthetic questions verified by a 12-consensus tool. It was authored by nguyen599 and last updated on Hugging Face in May 2026.
Use Cases
Analyze common failure modes in math synthesis based on the 520k error traces.
Train models to avoid systematic errors based on the documented error traces.
Benchmark model performance on synthetic math questions based on the updated datapoints.
Improve synthetic data generation quality based on the 12-consensus verification method described.
Strengths
Contains 520,000 error traces for detailed failure analysis.
Includes 50,000 new high-quality synthetic datapoints verified by a 12-consensus tool.
Actively maintained with updates documented in March 2026.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count for the primary dataset is unknown, which may limit suitability assessment.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
huggingface
Collection Method
Synthetic generation and error collection during model synthesis processes.
Freshness
Last updated 2026-05-09 09:04:11; freshness should be verified.
License is unknown; terms of use must be verified before application.