# Accuracy Improvement with SuperCorrect Across Topics
## Chart Type
Bar chart comparing accuracy improvements between Base LLM and SuperCorrect across mathematical topics.
## Axes
- **X-axis**: Accuracy (ranging from 0.0 to 0.8)
- **Y-axis**: Mathematical topics (listed vertically):
1. Algebra
2. Counting & Probability
3. Geometry
4. Intermediate Algebra
5. Number Theory
6. Prealgebra
7. Precalculus
## Legend
- **Location**: Top-right corner
- **Entries**:
- **Blue**: Base LLM
- **Green**: SuperCorrect Improvement
## Data Points
| Topic | Base LLM Accuracy | SuperCorrect Accuracy | Improvement (%) |
|------------------------|-------------------|-----------------------|-----------------|
| Algebra | 0.76 | 0.885 | +12.5% |
| Counting & Probability | 0.53 | 0.654 | +15.4% |
| Geometry | 0.44 | 0.557 | +11.7% |
| Intermediate Algebra | 0.36 | 0.471 | +21.0% |
| Number Theory | 0.51 | 0.726 | +21.5% |
| Prealgebra | 0.78 | 0.834 | +5.4% |
| Precalculus | 0.345 | 0.58 | +23.7% |
## Key Trends
1. **SuperCorrect consistently outperforms Base LLM** across all topics.
2. **Highest improvement**: Precalculus (+23.7%).
3. **Lowest improvement**: Prealgebra (+5.4%).
4. **Intermediate Algebra** shows the largest absolute accuracy gain (0.111).
5. **Number Theory** has the highest SuperCorrect accuracy (0.726).
## Spatial Grounding
- **Legend**: Top-right corner (blue = Base LLM, green = SuperCorrect).
- **Bars**: Horizontal, grouped by topic on the y-axis.
- **X-axis**: Accuracy values increase left-to-right.
## Component Isolation
1. **Header**: Title "Accuracy Improvement with SuperCorrect Across Topics".
2. **Main Chart**: Bar visualization with dual-color coding.
3. **Footer**: No additional text or elements.
## Verification
- All legend colors match bar colors (blue = Base LLM, green = SuperCorrect).
- Percentages align with calculated differences (e.g., Algebra: 0.885 - 0.76 = 0.125 = +12.5%).
- No missing or misaligned data points.
## Notes
- The chart uses **absolute percentage improvements** (difference in accuracy points), not relative gains.
- Precalculus shows the most significant jump due to its low Base LLM accuracy (0.345) and high SuperCorrect accuracy (0.58).
