Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

LLM Inference Performance on Synthetic Poisson Process Requests

18.7TPOT Speedup (%)

FlashSampling

-0.4364.5329.514.468Mar 16, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
18.7
2026.03
17.7
2026.03
15.2
2026.03
14.5
2026.03
12.8
2026.03
10.8
2026.03
9.7
2026.03
6.9
2026.03
5.2
2026.03
4.4
2026.03
3.7
2026.03
3.6
2026.03
3
2026.03
2.8
2026.03
2.4
2026.03
2.1
2026.03
1.9
2026.03
1.8
2026.03
1.8
2026.03
1.7
2026.03
1.6
2026.03
1.6
2026.03
1.5
2026.03
1.3
2026.03
1.2
2026.03
1.1
2026.03
1.1
2026.03
0.3