Creative Writing on Alpaca SFT Short Stories
3.7Self-BLEU (Diversity)λ-PR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| λ-PRBackbone=Mistral-12B2026.04 | 3.7 | 2.5 | |
| λ-PRBackbone=LLAMA-3.1-8B2026.04 | 3.8 | 2.7 | |
| λ-PRBackbone=OLMo-2-13B2026.04 | 3.9 | 3 | |
| λ-PRBackbone=PHI-4-14B2026.04 | 4 | 3.1 | |
| λ-PRBackbone=Pythia-12B2026.04 | 4.1 | 2.1 | |
| λ-PRBackbone=QWEN-3-8B2026.04 | 5 | 3.4 | |
| BASEBackbone=Pythia-12B2026.04 | 9.3 | 2.9 | |
| TOFUBackbone=Pythia-12B2026.04 | 10.8 | 3.6 | |
| TOFUBackbone=LLAMA-3.1-8B2026.04 | 11.2 | 4.5 | |
| BASEBackbone=Mistral-12B2026.04 | 11.4 | 3.9 | |
| BASEBackbone=LLAMA-3.1-8B2026.04 | 11.5 | 3.5 | |
| GEMBackbone=LLAMA-3.1-8B2026.04 | 11.5 | 4.5 | |
| TOFUBackbone=OLMo-2-13B2026.04 | 11.9 | 4.6 | |
| BASEBackbone=OLMo-2-13B2026.04 | 12.3 | 3.7 | |
| TOFUBackbone=PHI-4-14B2026.04 | 12.4 | 4.6 | |
| TOFUBackbone=Mistral-12B2026.04 | 12.7 | 4.6 | |
| TOFUBackbone=QWEN-3-8B2026.04 | 12.7 | 4.5 | |
| GEMBackbone=Pythia-12B2026.04 | 13 | 3.4 | |
| GEMBackbone=OLMo-2-13B2026.04 | 13.2 | 4.6 | |
| GEMBackbone=PHI-4-14B2026.04 | 13.4 | 4.7 | |
| BASEBackbone=PHI-4-14B2026.04 | 13.8 | 3.4 | |
| GEMBackbone=QWEN-3-8B2026.04 | 13.9 | 4.5 | |
| GEMBackbone=Mistral-12B2026.04 | 14.3 | 4.6 | |
| FLBackbone=LLAMA-3.1-8B2026.04 | 14.3 | 4.5 | |
| FLBackbone=Pythia-12B2026.04 | 14.6 | 3.8 | |
| FLBackbone=OLMo-2-13B2026.04 | 15.9 | 4.7 | |
| FLBackbone=PHI-4-14B2026.04 | 16.2 | 4.7 | |
| FLBackbone=QWEN-3-8B2026.04 | 16.4 | 4.6 | |
| FLBackbone=Mistral-12B2026.04 | 16.8 | 4.7 | |
| CEBackbone=QWEN-3-8B2026.04 | 22.3 | 4.6 | |
| CEBackbone=Mistral-12B2026.04 | 22.5 | 4.8 | |
| CEBackbone=LLAMA-3.1-8B2026.04 | 22.6 | 4.7 | |
| CEBackbone=Pythia-12B2026.04 | 23.8 | 3.7 | |
| CEBackbone=OLMo-2-13B2026.04 | 24.5 | 4.8 | |
| CEBackbone=PHI-4-14B2026.04 | 24.8 | 4.8 | |
| BASEBackbone=QWEN-3-8B2026.04 | 24.9 | 2.9 |