# Technical Document Extraction: Heatmap Analysis ## Title "Pressure Testing DeepSeek-V3 128K Context via 'Needle In A HayStack'" ## Axes and Labels - **X-Axis**: - Label: `Context Length (#Tokens)` - Markers: `2K`, `11K`, `20K`, `29K`, `38K`, `47K`, `56K`, `65K`, `74K`, `83K`, `92K`, `101K`, `110K`, `119K`, `128K` - Range: `2K` to `128K` (increasing in ~9K increments) - **Y-Axis**: - Label: `Document Depth Percent (%)` - Markers: `0`, `7`, `14`, `21`, `29`, `36`, `43`, `50`, `57`, `64`, `71`, `79`, `86`, `93`, `100` - Range: `0%` to `100%` (increasing in ~7% increments) - **Colorbar**: - Label: `Score` - Range: `1` (red) to `10` (green) - Gradient: Red → Yellow → Green (low to high scores) ## Chart Type and Observations - **Heatmap**: - Entire grid is uniformly colored in **bright green**, indicating **maximum scores (10)** across all data points. - No variation in color intensity observed, suggesting consistent performance metrics for all combinations of context length and document depth. ## Key Trends 1. **Score Consistency**: - All context lengths (2K–128K tokens) and document depths (0%–100%) achieve the highest possible score (`10`). - No degradation in performance at extreme values (e.g., 128K tokens or 100% depth). 2. **Color Interpretation**: - Green dominance confirms optimal performance under all tested conditions. - No red/yellow regions to indicate suboptimal scores. ## Technical Notes - The heatmap visualizes the relationship between **context length** (input size) and **document depth** (retrieval depth) for the DeepSeek-V3 model. - The uniform green color implies the model maintains high accuracy or relevance across all tested scenarios. - No explicit data table is present; the heatmap serves as the sole data representation. ## Conclusion The visualization demonstrates that DeepSeek-V3 128K context performs consistently at maximum capacity (`Score = 10`) regardless of input size or retrieval depth. This suggests robustness in handling large-scale context and deep document traversal.