## Dual-Axis Line Chart: Rule Set Accuracy and Emotion Purity Across Layers
### Overview
This is a dual-axis line chart tracking two distinct metrics across 6 sequential layers: **rule-based accuracy** (left y-axis) and **average purity of emotion detection** (right y-axis) for four emotions: Anger, Joy, Optimism, and Sadness. All metrics show a consistent upward trend as layer number increases.
### Components/Axes
- **X-axis**: Labeled *Layer*, with discrete markers at positions 1, 2, 3, 4, 5, 6 (bottom-center horizontal axis).
- **Left Y-axis**: Labeled *Accuracy*, with a linear scale ranging from 0.4 to 0.8, in increments of 0.1.
- **Right Y-axis**: Labeled *Average Purity*, with a linear scale ranging from 1.1 to 1.8, in increments of 0.1.
- **Legend (top-left corner, above chart area)**:
- Black solid line with square markers: *Accuracy - Rule Set*
- Blue dashed line with circle markers: *Purity - Anger*
- Orange dashed line with square markers: *Purity - Joy*
- Green dotted line with triangle markers: *Purity - Optimism*
- Red dotted line with triangle markers: *Purity - Sadness*
### Detailed Analysis
All data points are approximate values, with ±0.02 uncertainty for accuracy, ±0.05 uncertainty for purity:
1. **Accuracy - Rule Set (black solid line, left axis)**
- Trend: Steadily increasing, with a sharp upward jump between Layer 2 and 3, then continues rising to near the maximum scale value.
- Data points:
- Layer 1: ~0.39
- Layer 2: ~0.44
- Layer 3: ~0.63
- Layer 4: ~0.67
- Layer 5: ~0.79
- Layer 6: ~0.80
2. **Purity - Anger (blue dashed line, right axis)**
- Trend: Consistent upward slope, accelerating slightly after Layer 3, ending at the maximum purity scale value.
- Data points:
- Layer 1: ~1.10
- Layer 2: ~1.15
- Layer 3: ~1.35
- Layer 4: ~1.45
- Layer 5: ~1.70
- Layer 6: ~1.80
3. **Purity - Joy (orange dashed line, right axis)**
- Trend: Steady increase, with a sharp rise between Layer 2 and 3, matching Anger's final purity value.
- Data points:
- Layer 1: ~1.15
- Layer 2: ~1.20
- Layer 3: ~1.38
- Layer 4: ~1.50
- Layer 5: ~1.75
- Layer 6: ~1.80
4. **Purity - Optimism (green dotted line, right axis)**
- Trend: Gradual, slow increase, with the lowest final purity value among all emotion metrics.
- Data points:
- Layer 1: ~1.20
- Layer 2: ~1.25
- Layer 3: ~1.32
- Layer 4: ~1.40
- Layer 5: ~1.65
- Layer 6: ~1.70
5. **Purity - Sadness (red dotted line, right axis)**
- Trend: Slow initial growth, then a steeper rise after Layer 3, ending between Optimism and Anger/Joy.
- Data points:
- Layer 1: ~1.10
- Layer 2: ~1.12
- Layer 3: ~1.28
- Layer 4: ~1.35
- Layer 5: ~1.60
- Layer 6: ~1.75
### Key Observations
- All metrics (accuracy and all purity scores) increase monotonically with layer number (1 to 6).
- The *Accuracy - Rule Set* has the most dramatic growth, with a 43% increase between Layer 2 and 3.
- *Purity - Anger* and *Purity - Joy* reach the highest purity value (~1.80) at Layer 6, tying for the top position.
- *Purity - Optimism* has the lowest final purity (~1.70) and the most gradual growth trajectory.
- At Layer 1, all purity metrics cluster between 1.10-1.20, while accuracy starts at ~0.39.
### Interpretation
This chart demonstrates that deeper model layers (1 to 6) drive improvements in both rule-based task accuracy and emotion detection purity. The sharp accuracy jump between Layer 2 and 3 suggests a critical layer where the model learns key features for rule application, which correlates with a parallel rise in emotion purity. The fact that Anger and Joy reach the highest purity implies these emotions may be more distinct or easier for the model to isolate in deeper layers, while Optimism (with the lowest final purity) may be a more ambiguous or complex emotion for the model to detect with high precision. The consistent upward trend across all metrics indicates that deeper layers enhance both rule-based performance and refined emotion identification, suggesting a synergistic improvement in model capabilities as depth increases.
