Local Code Arena: MBPP Benchmark Results for Qwen 2.5 Coder 14B

Raw evaluation metrics and execution telemetry logs from running the Mostly Basic Python Problems (MBPP) benchmark against the Qwen 2.5 Coder 14B model. The dataset captures performance data from a local consumer hardware evaluation, establishing a heavy-parameter upper bound. It was authored by ShahzebKhoso and last updated on May 29, 2026.

Use Cases

• Analyzing model performance on basic Python problems based on the MBPP benchmark.
• Comparing execution telemetry and structural syntax outputs across different hardware configurations.
• Studying the interaction between model capacity and strict functional code correctness.

Strengths

• Captures telemetry from a specific heavyweight model (Qwen 2.5 Coder 14B) as an upper bound.
• Focuses on a standardized benchmark (MBPP) for reproducible evaluation.

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

ShahzebKhoso on Hugging Face

Collection Method

Execution of the MBPP benchmark against the Qwen 2.5 Coder 14B model on local consumer hardware.

Freshness

Last updated 2026-05-29 06:20:51; freshness should be verified.