compare-results

What problem does it solve?

This Skill helps users determine whether a quantized model checkpoint is acceptable by comparing it against a validated baseline and identifying meaningful accuracy changes.

Core Features & Use Cases

• Baseline vs Candidate Analysis: Establishes matching baseline and candidate checkpoints, verifies evaluation readiness, and computes reliable metric deltas.
• Evaluation Comparability Checks: Validates prompts, generation settings, datasets, sample counts, reasoning handling, judges, and score fields before interpreting results.
• Quantization Feasibility Decisions: Produces evidence-based feasibility verdicts for quantized checkpoints and avoids unsupported conclusions from incomplete comparisons.

Quick Start

Compare the baseline and quantized evaluation runs and determine whether the quantized checkpoint is acceptable.