Local Code Arena: MBPP Benchmark Results for Code Llama 7B

Local Code Arena Telemetry contains raw evaluation metrics, execution logs, and syntax outputs from running the Mostly Basic Python Problems (MBPP) benchmark against Meta's Code Llama 7B model. The dataset, hosted by ShahzebKhoso on Hugging Face, documents baseline performance for early-generation specialized code engines. It was last updated on 2026-05-24.

Use Cases

• Benchmarking code generation model performance based on MBPP evaluation metrics
• Analyzing execution telemetry logs for model behavior insights
• Comparing early-generation code engine outputs with modern models based on the described chronological anchor point
• Studying structural syntax outputs from code generation tasks

Strengths

• Contains raw evaluation metrics, execution telemetry logs, and structural syntax outputs as described
• Serves as a documented chronological anchor point for measuring post-training alignment improvements
• Focuses on a specific, widely-used benchmark (MBPP) and model (Code Llama 7B)

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 and file formats are unknown, which may limit suitability assessment

Provenance

Source

Hugging Face dataset by ShahzebKhoso

Collection Method

Results from running the MBPP benchmark against Meta's Code Llama 7B model

Freshness

Last updated 2026-05-24 11:10:24; freshness should be verified