Sonnet Opus 4.5 4.6 Gemini 3.0 3.1 Pro Gpt 5 5.1 5.2 Glm 4.7 Minimax M2.1 Deepseek V3.2 Hi

VINAY-UMRETHE's dataset aggregates 70.2K examples of high-quality outputs from frontier large language models. It is designed for training and distilling LLMs to exhibit advanced Chain-of-Thought, Agentic, Mathematical, and Coding capabilities. The data is formatted into a strict OpenAI messages structure and was last updated on June 9, 2026.

Use Cases

• Fine-tuning language models for Chain-of-Thought reasoning based on the described multi-source instruction data.
• Distilling advanced capabilities from frontier models into smaller models based on the aggregated high-quality outputs.
• Training models for agentic task execution based on the dataset's structured instruction-response format.
• Improving model performance on mathematical problem-solving tasks using the described reasoning examples.
• Enhancing code generation and understanding in LLMs based on the coding capability examples in the dataset.

Strengths

• Contains 70.2K total examples, providing a substantial volume of training data.
• Aggregates outputs from multiple frontier models, including Sonnet, Opus, Gemini, GPT, GLM, MiniMax, and DeepSeek.
• Structured in a strict OpenAI messages format, which may aid compatibility with common training pipelines.

Limitations

• Description metadata is limited; actual data quality, column definitions, and subset details require manual inspection after download.
• Row count for individual subsets is unknown, which may limit suitability assessment for specific training tasks.
• The dataset's source aggregation method may introduce biases inherent to the contributing models and their training data.

Provenance

Source

VINAY-UMRETHE on Hugging Face

Collection Method

Aggregation of high-quality outputs from multiple frontier large language models.

Freshness

Last updated 2026-06-09 02:12:59; freshness should be verified.