## Bar Chart: Model Accuracy Comparison
### Overview
The chart compares the accuracy of two models, **ProofWriter** and **LogicNLI**, under four configurations: **"Ours"**, **"w/o Decomposer"**, **"w/o Search Router"**, and **"w/o Resolver"**. Accuracy is measured in percentage, with values ranging from 30% to 90%.
### Components/Axes
- **X-axis**: Model names (**ProofWriter**, **LogicNLI**).
- **Y-axis**: Accuracy (%) from 30% to 90%.
- **Legend**:
- **Dark Red**: "Ours"
- **Medium Red**: "w/o Decomposer"
- **Light Red**: "w/o Search Router"
- **Light Pink**: "w/o Resolver"
### Detailed Analysis
#### ProofWriter
- **"Ours"**: 88.5% (highest accuracy, dark red bar).
- **"w/o Decomposer"**: 62.0% (medium red bar, ~26.5% drop from "Ours").
- **"w/o Search Router"**: 37.7% (light red bar, ~50.8% drop from "Ours").
- **"w/o Resolver"**: 47.1% (light pink bar, ~47.4% drop from "Ours").
#### LogicNLI
- **"Ours"**: 70.7% (dark red bar).
- **"w/o Decomposer"**: 42.8% (medium red bar, ~39.8% drop from "Ours").
- **"w/o Search Router"**: 39.1% (light red bar, ~44.9% drop from "Ours").
- **"w/o Resolver"**: 57.5% (light pink bar, ~18.8% drop from "Ours").
### Key Observations
1. **"Ours"** consistently achieves the highest accuracy for both models.
2. Removing the **Decomposer** causes the largest accuracy drop in **ProofWriter** (26.5%) compared to **LogicNLI** (39.8%).
3. The **Search Router** has a more severe impact on **ProofWriter** (50.8% drop) than on **LogicNLI** (44.9% drop).
4. The **Resolver** has the smallest impact on **LogicNLI** (18.8% drop) but a moderate impact on **ProofWriter** (47.4% drop).
### Interpretation
- The **"Ours"** configuration demonstrates the highest performance, indicating that all components (Decomposer, Search Router, Resolver) are critical for optimal accuracy.
- The **Decomposer** is particularly vital for **ProofWriter**, as its absence leads to a significant performance decline.
- The **Search Router** appears more critical for **ProofWriter**, while the **Resolver** is more impactful for **LogicNLI**.
- The data suggests that component importance varies by model, highlighting the need for tailored optimization strategies.
### Spatial Grounding & Trend Verification
- **Legend Position**: Bottom of the chart, with colors matching bar colors exactly.
- **Bar Order**: For each model, bars are ordered left-to-right as "Ours", "w/o Decomposer", "w/o Search Router", "w/o Resolver".
- **Trend Check**:
- **ProofWriter**: "Ours" > "w/o Decomposer" > "w/o Resolver" > "w/o Search Router".
- **LogicNLI**: "Ours" > "w/o Resolver" > "w/o Decomposer" > "w/o Search Router".
- **Outliers**: The drastic drop in **ProofWriter** without the **Search Router** (37.7%) is an anomaly compared to other configurations.
### Content Details
- All values are explicitly labeled on the bars (e.g., 88.5%, 62.0%).
- No additional text or hidden data is present.
- Colors are consistent with the legend (e.g., dark red = "Ours").
