Generative Tasks on 8-Task Generative Suite
100AccuracyHF FP16
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HF FP16Model=1B, Scenario=unconstrained2026.04 | 100 | — | |
| NVE BaseModel=3B, Scenario=unconstrained2026.04 | 100 | 2.12 | |
| llama.cppModel=3B, Scenario=unconstrained2026.04 | 100 | — | |
| NVE BaseModel=3B, Scenario=2 GB2026.04 | 100 | 2.31 | |
| llama.cppModel=8B, Scenario=unconstrained2026.04 | 100 | — | |
| llama.cppModel=8B, Scenario=8 GB2026.04 | 100 | — | |
| Llama-3.2-3BQuantization Strategy=W4A16, n=82026.04 | 100 | — | |
| Llama-3.2-3BQuantization Strategy=W4A8, n=82026.04 | 100 | — | |
| Llama-3.2-3BQuantization Strategy=MCAP Mixed, n=82026.04 | 100 | — | |
| Llama-3.1-8BQuantization Strategy=W4A16, n=82026.04 | 100 | — | |
| Llama-3.1-8BQuantization Strategy=W4A8, n=82026.04 | 100 | — | |
| NVE BaseModel=1B, Scenario=unconstrained2026.04 | 88 | 3.22 | |
| llama.cppModel=1B, Scenario=unconstrained2026.04 | 88 | — | |
| NVE BaseModel=1B, Scenario=2 GB2026.04 | 88 | 4.23 | |
| NVE BaseModel=8B, Scenario=unconstrained2026.04 | 88 | 0.79 | |
| HF FP16Model=8B, Scenario=unconstrained2026.04 | 88 | — | |
| NVE BaseModel=8B, Scenario=4 GB2026.04 | 88 | 1.25 | |
| NVE BaseModel=8B, Scenario=8 GB2026.04 | 88 | 0.55 | |
| Llama-3.2-1BQuantization Strategy=W4A16, n=82026.04 | 88 | — | |
| Llama-3.2-1BQuantization Strategy=W4A8, n=82026.04 | 88 | — | |
| Llama-3.2-1BQuantization Strategy=MCAP Mixed, n=82026.04 | 88 | — |