Diversity Measurement on NoveltyBench curated
48.1D_Can MeanOLMo-2-7B Base
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| OLMo-2-7B BaseStage=Base, n=392026.06 | 48.1 | 0.03 | — | — | — | |
| OLMo-2-7B SFTStage=SFT, n=392026.06 | 36.9 | 0.085 | — | — | — | |
| OLMo-2-7B DPOStage=DPO, n=392026.06 | 31.2 | 0.082 | — | — | — | |
| OLMo-2-7B Instruct (RLVR)Stage=Instruct (RLVR), n=392026.06 | 30.3 | 0.086 | — | — | — | |
| Base > Instruct (RLVR) ContrastContrast=Base>Instruct (RLVR)2026.06 | — | — | 0.179 | 1.889 | 2.3 | |
| Base > SFT ContrastContrast=Base>SFT2026.06 | — | — | 0.112 | 1.212 | 4.1 | |
| DPO ≠ Instruct (H1') ContrastContrast=DPO ≠ Instruct (H1')2026.06 | — | — | 0.01 | 0.375 | 0.028 | |
| SFT > DPO ContrastContrast=SFT>DPO2026.06 | — | — | 0.057 | 0.957 | 2.8 |