This image is a 2D line chart displaying the relationship between "JSD" (Y-axis) and "Token Position" (X-axis) for three different methods, each represented by a distinct colored line and a shaded confidence interval. ### Chart Components: 1. **X-axis:** * **Label:** "Token Position" * **Range:** From 0 to approximately 24-25. * **Major Tick Markers:** 0, 10, 20. * **Minor Tick Markers:** Implicitly at intervals of 1, though only major ticks are labeled. 2. **Y-axis:** * **Label:** "JSD" * **Range:** From 0.0 to 0.7. * **Major Tick Markers:** 0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7. 3. **Legend:** * **Placement:** Located in the bottom-right quadrant of the plot area. * **Entries:** * A dark blue line represents "REAP". * A light blue line represents "M-SMoE". * A yellow/gold line represents "HC-SMoE". ### Data Series and Trends: All three data series start at approximately (0, 0.0) and generally show an increasing trend in JSD as Token Position increases. Each line is accompanied by a shaded area, indicating a confidence interval around the mean performance. 1. **REAP (Dark Blue Line):** * **Trend:** The dark blue line representing REAP starts at JSD 0.0 at Token Position 0. It rises sharply initially, reaching approximately JSD 0.32 by Token Position 2. It then shows a more gradual, but consistent, upward trend with some minor fluctuations. The slope generally decreases over higher token positions. * **Approximate Data Points:** * (0, 0.0) * (2, 0.32) * (5, 0.32) * (10, 0.38) * (15, 0.48) * (20, 0.53) * (24, 0.62) * **Confidence Interval:** The shaded area around the REAP line is a light blue-grey. 2. **M-SMoE (Light Blue Line):** * **Trend:** The light blue line representing M-SMoE also starts at JSD 0.0 at Token Position 0. It exhibits the steepest initial rise among all methods, reaching approximately JSD 0.48 by Token Position 2. It generally maintains the highest JSD values across most of the Token Positions, showing a continuous upward trend with a decreasing slope. * **Approximate Data Points:** * (0, 0.0) * (2, 0.48) * (5, 0.53) * (10, 0.57) * (15, 0.60) * (20, 0.62) * (24, 0.65) * **Confidence Interval:** The shaded area around the M-SMoE line is a very light blue. 3. **HC-SMoE (Yellow/Gold Line):** * **Trend:** The yellow/gold line representing HC-SMoE begins at JSD 0.0 at Token Position 0. It shows a strong initial increase, reaching approximately JSD 0.38 by Token Position 2. For most of the chart, HC-SMoE performs better than REAP but slightly below M-SMoE. However, towards the end of the chart, its performance converges with and slightly surpasses M-SMoE. * **Approximate Data Points:** * (0, 0.0) * (2, 0.38) * (5, 0.43) * (10, 0.52) * (15, 0.57) * (20, 0.61) * (24, 0.66) * **Confidence Interval:** The shaded area around the HC-SMoE line is a light yellow/beige. ### Summary of Performance: * All three methods show an improvement in JSD as Token Position increases. * M-SMoE generally achieves the highest JSD values for most Token Positions, particularly in the early and middle ranges. * HC-SMoE consistently outperforms REAP and shows a strong performance, eventually surpassing M-SMoE at the highest Token Positions (around 24). * REAP consistently shows the lowest JSD values among the three methods across the entire range of Token Positions. * The confidence intervals indicate the variability of the JSD values for each method at different token positions.

# Technical Data Extraction: JSD vs. Token Position Chart ## 1. General Description This image is a line graph illustrating the relationship between **Token Position** (x-axis) and **JSD** (Jensen-Shannon Divergence, y-axis) for three different models or methods. Each method is represented by a solid colored line accompanied by a semi-transparent shaded area indicating a confidence interval or variance. ## 2. Axis Information * **Y-Axis Label:** JSD * **Y-Axis Scale:** 0.0 to 0.7, with major tick marks every 0.1 units. * **X-Axis Label:** Token Position * **X-Axis Scale:** 0 to 25, with major tick marks every 10 units (0, 10, 20). The data points appear to be sampled at integer intervals from 0 to 25. ## 3. Legend and Data Series The legend is located in the bottom-right quadrant of the plot: | Series Label | Line Color | Shaded Area Color | | :--- | :--- | :--- | | **REAP** | Dark Blue | Light Blue-Grey | | **M-SMoE** | Cyan / Light Blue | Very Light Cyan | | **HC-SMoE** | Gold / Olive Yellow | Light Yellow-Green | ## 4. Key Trends and Data Points All three series show a positive correlation, where JSD increases as the Token Position increases, eventually showing signs of plateauing or slowing growth toward the end of the sequence. ### REAP (Dark Blue) * **Initial State:** Starts at (0, 0). * **Growth:** Shows the slowest initial growth compared to the other two models. * **Mid-point:** Reaches a JSD of approximately 0.4 at Token Position 10. * **Final State:** Ends at a JSD of approximately 0.6 at Token Position 25. * **Variance:** Maintains a relatively consistent shaded variance band throughout the sequence. ### M-SMoE (Cyan) * **Initial State:** Starts at (0, 0). * **Growth:** Exhibits the sharpest initial increase, jumping to ~0.45 by Token Position 2. * **Mid-point:** Reaches a JSD of approximately 0.55–0.58 at Token Position 10. * **Final State:** Ends at a JSD of approximately 0.63 at Token Position 25. * **Trend:** This model maintains the highest JSD values for the majority of the sequence (from position 1 to 20). ### HC-SMoE (Gold) * **Initial State:** Starts at (0, 0). * **Growth:** Growth rate is intermediate between REAP and M-SMoE. * **Mid-point:** Reaches a JSD of approximately 0.5 at Token Position 10. * **Final State:** Ends at the highest final JSD value of approximately 0.65 at Token Position 25. * **Trend:** While it starts lower than M-SMoE, it converges with and slightly overtakes M-SMoE around Token Position 22. ## 5. Comparative Analysis * **Early Sequence (0-5):** M-SMoE has the highest JSD, followed by HC-SMoE, with REAP being the lowest. * **Late Sequence (20-25):** HC-SMoE and M-SMoE converge at the top of the scale (0.6-0.65), while REAP remains the lowest (0.55-0.6). * **Uncertainty:** The shaded regions for all three models overlap significantly after Token Position 15, suggesting that the performance differences between the models become less statistically distinct as the sequence length increases.

## Line Chart: JSD vs Token Position ### Overview The line chart displays the JSD (Joint Similarity Distribution) values across different token positions. The chart shows three different methods: REAP, M-SMoE, and HC-SMoE, each represented by a distinct line color. ### Components/Axes - **X-axis (Token Position)**: The horizontal axis represents the token positions, ranging from 0 to 20. - **Y-axis (JSD)**: The vertical axis represents the JSD values, ranging from 0.0 to 0.7. - **Legend**: The legend on the right side of the chart identifies the three methods used to calculate the JSD. ### Detailed Analysis or ### Content Details - **REAP**: The blue line shows the highest JSD values, indicating the most similarity between tokens at different positions. - **M-SMoE**: The light blue line has moderate JSD values, suggesting a balance between similarity and dissimilarity. - **HC-SMoE**: The yellow line has the lowest JSD values, indicating the least similarity between tokens at different positions. ### Key Observations - All three methods show an increasing trend in JSD values as the token position increases. - The REAP method consistently has the highest JSD values, while the HC-SMoE method has the lowest. - The M-SMoE method shows a more gradual increase in JSD values compared to the other two methods. ### Interpretation The data suggests that the REAP method captures more similarity between tokens across different positions compared to the M-SMoE and HC-SMoE methods. This could indicate that the REAP method is more effective at identifying and preserving the relationships between tokens in the dataset. The M-SMoE method, while still capturing similarity, does so at a more moderate rate, and the HC-SMoE method, which has the lowest JSD values, may be less effective at preserving the relationships between tokens.

## Chart Description This is a line chart plotting **JSD** on the y-axis against **Token Position** on the x-axis. The chart compares three different models: **REAP**, **M-SMoE**, and **HC-SMoE**. Each model is represented by a colored line with a corresponding shaded area indicating the confidence interval or standard deviation. ### Axes * **Y-axis:** Labeled "JSD". The scale ranges from 0.0 to 0.7, with major tick marks at 0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, and 0.7. * **X-axis:** Labeled "Token Position". The scale ranges from 0 to 25, with major tick marks at 0, 10, and 20. ### Legend The legend is located in the bottom right corner of the chart and defines the colors for each model: * **REAP:** Dark blue line. * **M-SMoE:** Light blue line. * **HC-SMoE:** Gold/yellow line. ### Data Trends All three models show an increasing trend in JSD as the Token Position increases, starting from a JSD of 0.0 at Token Position 0. 1. **REAP (Dark Blue):** * The line starts at (0, 0) and rises steeply to approximately 0.3 at Token Position 2. * The increase becomes more gradual, reaching a JSD of approximately 0.6 at Token Position 25. * The shaded area around the line indicates a moderate level of variability, roughly ±0.05 JSD units. 2. **M-SMoE (Light Blue):** * The line starts at (0, 0) and shows the steepest initial rise among the three models, reaching approximately 0.45 at Token Position 2. * It continues to rise, reaching a JSD of approximately 0.62 at Token Position 25. * The shaded area is wider than REAP's, indicating higher variability, roughly ±0.05-0.08 JSD units. 3. **HC-SMoE (Gold/Yellow):** * The line starts at (0, 0) and rises steeply to approximately 0.35 at Token Position 2. * It continues to rise, generally staying between the REAP and M-SMoE lines, and reaches the highest JSD of approximately 0.65 at Token Position 25. * The shaded area is similar in width to M-SMoE's, indicating comparable variability. Overall, M-SMoE and HC-SMoE show a faster initial increase in JSD compared to REAP. HC-SMoE achieves the highest final JSD value at Token Position 25, followed by M-SMoE and then REAP.

# Technical Document Extraction: Line Chart Analysis ## Chart Type Line chart with shaded confidence intervals. ## Axis Labels - **X-axis**: Token Position (0 to 20, increments of 10) - **Y-axis**: JSD (0.0 to 0.7, increments of 0.1) ## Legend - **REAP**: Dark blue line - **M-SMoE**: Light blue line - **HC-SMoE**: Gold line ## Key Trends 1. **Initial Behavior**: - All lines start at (0, 0) and rise sharply in the first 5 token positions. - HC-SMoE (gold) begins highest, followed by M-SMoE (light blue), then REAP (dark blue). 2. **Mid-Range (Token Positions 5–15)**: - Lines diverge further, with HC-SMoE maintaining the highest JSD values. - M-SMoE shows moderate growth, while REAP lags slightly behind. 3. **Final Segment (Token Positions 15–20)**: - All lines plateau near 0.6–0.65 JSD. - HC-SMoE remains the highest, with REAP approaching M-SMoE but not surpassing it. ## Shaded Regions - Gray bands around each line represent confidence intervals or error margins. - HC-SMoE exhibits the widest variability (broadest shaded region), particularly after token position 10. ## Data Points - **Token Position 0**: All lines at 0.0 JSD. - **Token Position 10**: - REAP: ~0.4 - M-SMoE: ~0.55 - HC-SMoE: ~0.5 - **Token Position 20**: - REAP: ~0.6 - M-SMoE: ~0.62 - HC-SMoE: ~0.65 ## Observations - Lines do not intersect; relative order (HC-SMoE > M-SMoE > REAP) is preserved throughout. - JSD values for all models increase monotonically with token position. - HC-SMoE demonstrates the steepest initial growth and highest final JSD.