Time-to-First-Token (TTFT) Latency on Controlled synthetic corpus Config B 1.0 (test)
40.1TTFT p50APC+Optimized
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| APC+OptimizedHardware=RTX 4090, Model=Qwen2.5-7B, Strategy=APC with optimized ordering (CacheWeaver)2026.06 | 40.1 | 117.4 | 51.4 | 19 | |
| APC+RetrievalHardware=RTX 4090, Model=Qwen2.5-7B, Strategy=APC with retrieval order2026.06 | 59.6 | 117.4 | 55.9 | 12.3 | |
| APC+SortedHardware=RTX 4090, Model=Qwen2.5-7B, Strategy=APC with lexicographic sorting2026.06 | 81.2 | 118.7 | 76.3 | 25.6 | |
| No CacheHardware=RTX 4090, Model=Qwen2.5-7B, Strategy=Baseline2026.06 | 121.3 | 123.7 | 121.4 | 0 |