Efficiency Measurement on Long-form literature review workload
22.4End-to-end Latency (s)Lazy-Attn
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Lazy-AttnModel=Tulu3-Block-FT 8B, Hardware=H1002026.06 | 22.4 | 1.7 | |
| Standard RAGModel=Tulu3-Block-FT 8B, Hardware=H1002026.06 | 38 | 1 |