Local Code Arena Starcoder2 15B: MBPP Benchmark Telemetry

ShahzebKhoso hosts raw evaluation metrics, execution telemetry logs, and structural syntax outputs from running the Mostly Basic Python Problems (MBPP) benchmark against the StarCoder2 15B base model. The dataset captures telemetry from conversational evaluation loops to establish a baseline for unaligned foundational weights. It was last updated on May 28, 2026.

Use Cases

• Benchmarking code generation models based on MBPP problem-solving performance.
• Analyzing execution telemetry logs to understand model behavior during evaluation.
• Studying structural syntax outputs for error patterns in generated code.
• Establishing baseline performance for unaligned foundational language model weights.

Strengths

• Focuses on a specific, widely-used benchmark (MBPP) for Python code generation.
• Captures multiple telemetry data types: raw metrics, execution logs, and syntax outputs.
• Provides a baseline for the StarCoder2 15B model, a known foundational model.

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 is unknown, which may limit suitability assessment.

Provenance

Source

Author ShahzebKhoso on Hugging Face.

Collection Method

Telemetry captured from running the MBPP benchmark against the StarCoder2 15B model.

Freshness

Last updated 2026-05-28 14:27:28; freshness should be verified.