# MiniCPM5-1B PX-Mod Evaluation Report

## Executive Summary

Comparison of MiniCPM5-1B Base vs PX-Patched (Peak and Subjective modes).

### Accuracy Summary

| Model | Overall Accuracy | Logic | Arithmetic | HLE |
|-------|------------------|-------|------------|-----|
| minicpm5-1b-base | 40.00% | 36.67% | 30.00% | 0.00% |
| minicpm5-1b-px | 41.67% | 40.00% | 30.00% | 0.00% |
| minicpm5-1b-px-subj | 38.33% | 36.67% | 30.00% | 0.00% |

## Qualitative Analysis (Creative & Synthesis)

### minicpm5-1b-base

- Average response length: 836.7 chars
- Average Phi (Stability): 1.0000

### minicpm5-1b-px

- Average response length: 878.3 chars
- Average Phi (Stability): 0.9654

### minicpm5-1b-px-subj

- Average response length: 862.1 chars
- Average Phi (Stability): 0.9589