UltraData-SFT-2605: Core-Domain SFT Data for MiniCPM5-1B Post-Training

Over 15 million data points covering math, code, knowledge, and instruction following form the full set of core-domain SFT data used for post-training the MiniCPM5-1B-SFT model. This dataset is a key representative of L3 refined data within the UltraData L0-L4 tiered data management framework. It was authored by openbmb and last updated on Hugging Face in May 2026.

Use Cases

• Instruction tuning of language models based on the described instruction-following data.
• Improving mathematical reasoning in models based on the math domain data.
• Enhancing code generation capabilities based on the code domain data.
• Knowledge injection for language models based on the knowledge domain data.
• Benchmarking data refinement techniques based on its position in the L0-L4 tiered framework.

Strengths

• Contains over 15 million data points, indicating a substantial scale.
• Covers multiple core domains for AI training: math, code, knowledge, and instruction following.
• Represents a defined level (L3 refined) within a structured data management framework.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

openbmb on Hugging Face.

Collection Method

Likely collected and refined for the post-training of the MiniCPM5-1B-SFT model.

Freshness

Last updated 2026-05-28 10:22:10; freshness should be verified.