# Technical Document Extraction: Bar Chart Analysis
## Image Description
The image is a **horizontal bar chart** comparing the **minimum score at the last step (%)** for various AI models. The chart uses **blue bars** to represent percentages, with numerical values explicitly labeled next to each bar. The y-axis lists model names, and the x-axis represents the percentage scale (0–60%).
---
### Key Components
1. **Axes**:
- **X-axis**: Labeled *"Minimum Score at Last Step (%)"*, ranging from 0 to 60%.
- **Y-axis**: Lists AI model names in descending order of performance (top to bottom).
2. **Legend**:
- No explicit legend is visible. All bars are uniformly colored **blue**, implying a single data series.
3. **Data Points**:
- Each bar corresponds to a model and its minimum score percentage. Values are explicitly labeled next to the bars.
---
### Data Extraction
| Model Name | Minimum Score (%) |
|-------------------------------------|-------------------|
| EurusPRM-Stage1 | 54.6 |
| EurusPRM-Stage2 | 52.9 |
| Math-Shepherd-PRM-7B | 44.5 |
| Skywork-PRM-7B | 42.2 |
| Skywork-PRM-1.5B | 30.9 |
| Qwen2.5-Math-7B-PRM800K | 26.8 |
| Qwen2.5-Math-PRM-72B | 18.0 |
| Qwen2.5-Math-PRM-7B | 17.5 |
| RLFHFlow-PRM-Deepseek-8B | 17.3 |
| Qwen2.5-Math-7B-Math-Shepherd | 9.8 |
| RLFHFlow-PRM-Mistral-8B | 9.1 |
---
### Trends and Observations
1. **Performance Ranking**:
- **Top Performers**: *EurusPRM-Stage1* (54.6%) and *EurusPRM-Stage2* (52.9%) dominate, achieving the highest minimum scores.
- **Mid-Range Models**: *Math-Shepherd-PRM-7B* (44.5%) and *Skywork-PRM-7B* (42.2%) follow closely.
- **Lower Performers**: *Qwen2.5-Math-7B-Math-Shepherd* (9.8%) and *RLFHFlow-PRM-Mistral-8B* (9.1%) show significantly lower scores.
2. **Visual Trend**:
- Bars decrease in length from top to bottom, indicating a **descending performance trend** across models.
3. **Notable Outliers**:
- *EurusPRM-Stage1* and *EurusPRM-Stage2* outperform all other models by a margin of ~10–15%.
- *RLFHFlow-PRM-Mistral-8B* has the lowest score (9.1%), nearly half of the top performers.
---
### Spatial Grounding
- **Legend**: Not explicitly visible. All bars use **blue**, suggesting a single category.
- **Data Point Colors**: All bars match the implied legend (blue = minimum score %).
---
### Conclusion
The chart highlights significant performance disparities among AI models, with *EurusPRM* variants leading and *RLFHFlow-PRM-Mistral-8B* lagging. No textual or diagrammatic elements beyond the bar chart are present.
