Multi-turn Dialogue on MT-Bench (Token/s and Speedup)
270.96Tokens/sEVICT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EVICTTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05 | 270.96 | — | |
| EVICTTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05 | 254.83 | — | |
| EAGLE-3Temperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05 | 221.97 | — | |
| EAGLE-3Temperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05 | 213.12 | — | |
| DDDTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05 | 212.33 | — | |
| DDDTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05 | 203.17 | — | |
| EAGLE-3 + EVICTTemperature=0, Backbone=Qwen3-30B-A3B2026.05 | 190.34 | — | |
| EAGLE-3 + EVICTTemperature=1, Backbone=Qwen3-30B-A3B2026.05 | 180.18 | — | |
| VanillaTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05 | 174.57 | — | |
| EAGLE-3Temperature=0, Backbone=Qwen3-30B-A3B2026.05 | 171.08 | — | |
| VanillaTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05 | 170.29 | — | |
| EAGLE-3Temperature=1, Backbone=Qwen3-30B-A3B2026.05 | 162.26 | — | |
| EAGLE-3 + DDDTemperature=0, Backbone=Qwen3-30B-A3B2026.05 | 159.16 | — | |
| EAGLE-3 + DDDTemperature=1, Backbone=Qwen3-30B-A3B2026.05 | 152.31 | — | |
| VanillaTemperature=0, Backbone=Qwen3-30B-A3B2026.05 | 143.1 | — | |
| VanillaTemperature=1, Backbone=Qwen3-30B-A3B2026.05 | 140.13 | — | |
| LookaheadTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05 | 133.59 | — | |
| LookaheadTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05 | 129.01 | — | |
| LookaheadTemperature=0, Backbone=Qwen3-30B-A3B2026.05 | 101.59 | — | |
| LookaheadTemperature=1, Backbone=Qwen3-30B-A3B2026.05 | 99 | — |