Multi-session Retrieval-Augmented Generation on NarrativeQA (test)
38.4F1 ScoreContextPilot
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ContextPilotModel=Llama3.3-70B-Instruct2025.11 | 38.4 | 18,468.8 | |
| LMCacheModel=Llama3.3-70B-Instruct2025.11 | 37.8 | 12,575.7 | |
| Radix CacheModel=Llama3.3-70B-Instruct2025.11 | 37.8 | 12,644.9 | |
| CacheBlendModel=Llama3.3-70B-Instruct2025.11 | 31.3 | 13,710.5 | |
| ContextPilotModel=Qwen3-32B2025.11 | 29.6 | 22,780.4 | |
| LMCacheModel=Qwen3-32B2025.11 | 28.4 | 15,514 | |
| Radix CacheModel=Qwen3-32B2025.11 | 28.4 | 15,598.8 | |
| CacheBlendModel=Qwen3-32B2025.11 | 19.8 | 16,913.2 | |
| ContextPilotModel=Qwen3-4B-Instruct-25072025.11 | 17.3 | 57,681.3 | |
| LMCacheModel=Qwen3-4B-Instruct-25072025.11 | 16 | 39,276.4 | |
| Radix CacheModel=Qwen3-4B-Instruct-25072025.11 | 16 | 39,492.4 | |
| CacheBlendModel=Qwen3-4B-Instruct-25072025.11 | 11.3 | 42,819.5 |