DeepSeek-ProverBench is a dataset for training and evaluating large language models on formal theorem proving in the Lean 4 environment. It was created by deepseek-ai using a recursive theorem proving pipeline powered by the DeepSeek-V3 model to decompose complex problems into subgoals. The dataset was last updated on April 30,我们发现了一个问题，输入中的描述是中文的，但输出要求是英文。根据指令，我需要将输入翻译成英文。让我重新处理。2025.
Use Cases
- Fine-tuning language models for formal theorem proving based on the described Lean 4 proof data.
- Benchmarking the reasoning capabilities of AI models based on the described theorem decomposition pipeline.
- Studying automated proof generation strategies based on the described subgoal resolution process.
Strengths
- Data was generated using a sophisticated pipeline powered by the DeepSeek-V3 model.
- Specifically designed for the modern Lean 4 theorem proving environment.
- Last updated on 2025-04-30, indicating recent maintenance.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count, file formats, and license are unknown, which may limit suitability assessment.
Provenance
- Source
- deepseek-ai via Hugging Face.
- Collection Method
- Data collected through a recursive theorem proving pipeline powered by DeepSeek-V3.
- Time Range
- null
- Freshness
- Last updated 2025-04-30 14:54:40.
- Geography
- null