Question Answering on TruthfulQA (Factuality, TTR)
57.89FactualityMin-k + VCM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Min-k + VCMSampler=Min-k, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 57.89 | 78.17 | |
| η-sampling + VCMSampler=η-sampling, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 57.28 | 78.85 | |
| Top-p + VCMSampler=Top-p, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 57.16 | 78.85 | |
| p-less + VCMSampler=p-less, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 57.16 | 79.35 | |
| ϵ-sampling + VCMSampler=ϵ-sampling, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 57.04 | 78.85 | |
| Top-k + VCMSampler=Top-k, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.79 | 79.32 | |
| Top-nσ + VCMSampler=Top-nσ, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.79 | 78.53 | |
| Typical + VCMSampler=Typical, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.43 | 79.12 | |
| Min-p + VCMSampler=Min-p, VCM=Enabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.43 | 78.8 | |
| p-lessSampler=p-less, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.3 | 76.6 | |
| TypicalSampler=Typical, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.06 | 76.88 | |
| η-samplingSampler=η-sampling, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 56.06 | 77.04 | |
| ϵ-samplingSampler=ϵ-sampling, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 55.94 | 77.04 | |
| Top-pSampler=Top-p, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 55.81 | 77.04 | |
| Top-kSampler=Top-k, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 55.57 | 77.59 | |
| Min-kSampler=Min-k, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 55.57 | 77.09 | |
| Min-pSampler=Min-p, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 54.59 | 77.11 | |
| Top-nσSampler=Top-nσ, VCM=Disabled, Base Model=LLaMA-3-8B-Instruct2026.06 | 54.47 | 77.06 |