Local Code Arena MBPP: Benchmark Telemetry for DeepSeek Coder 6.7B

Raw evaluation metrics and execution telemetry logs from running the Mostly Basic Python Problems (MBPP) benchmark against the DeepSeek Coder 6.7B model. This dataset, authored by ShahzebKhoso, provides a specific anchor point for evaluating generational alignment improvements in code-generation architectures. It was last updated on May 24, 2026.

Use Cases

• Analyze model performance on basic Python tasks based on the raw evaluation metrics.
• Study execution patterns and errors based on the telemetry logs.
• Compare legacy code-specialist model dynamics with newer architectures based on the described anchor point.
• Investigate structural syntax outputs from a 6.7B parameter model.

Strengths

• Focuses on a specific, widely-used benchmark (MBPP) for code generation.
• Captures detailed telemetry logs beyond simple pass/fail metrics.
• Provides a defined anchor point for architectural comparisons as mentioned in the description.

Limitations

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

Provenance

Source

huggingface

Collection Method

Telemetry captured from running the MBPP benchmark on the DeepSeek Coder 6.7B model.

Freshness

Last updated 2026-05-24 10:13:26; freshness should be verified.