Instruction Following on Alpaca SFT Small Prompts
46.6Self-BLEUCE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CEBackbone=PHI-4-14B2026.04 | 46.6 | 4.2 | |
| CEBackbone=OLMo-2-13B2026.04 | 45 | 4.3 | |
| CEBackbone=QWEN-3-8B2026.04 | 44.9 | 4.1 | |
| CEBackbone=Mistral-12B2026.04 | 44.5 | 4.2 | |
| CEBackbone=LLAMA-3.1-8B2026.04 | 44.5 | 4.2 | |
| CEBackbone=Pythia-12B2026.04 | 35.9 | 3.8 | |
| BASEBackbone=QWEN-3-8B2026.04 | 31.5 | 3.8 | |
| GEMBackbone=QWEN-3-8B2026.04 | 29.5 | 4 | |
| FLBackbone=Mistral-12B2026.04 | 29.3 | 4.1 | |
| FLBackbone=QWEN-3-8B2026.04 | 29.2 | 4 | |
| FLBackbone=OLMo-2-13B2026.04 | 29 | 4.2 | |
| FLBackbone=PHI-4-14B2026.04 | 28.9 | 4.2 | |
| FLBackbone=LLAMA-3.1-8B2026.04 | 28.3 | 4.1 | |
| GEMBackbone=Mistral-12B2026.04 | 27.7 | 4.1 | |
| GEMBackbone=PHI-4-14B2026.04 | 27.5 | 4.1 | |
| GEMBackbone=OLMo-2-13B2026.04 | 27 | 4.1 | |
| GEMBackbone=LLAMA-3.1-8B2026.04 | 25.9 | 4.1 | |
| TOFUBackbone=QWEN-3-8B2026.04 | 24.1 | 4 | |
| FLBackbone=Pythia-12B2026.04 | 22.2 | 3.6 | |
| TOFUBackbone=PHI-4-14B2026.04 | 21.8 | 4.1 | |
| TOFUBackbone=OLMo-2-13B2026.04 | 21.3 | 4.1 | |
| TOFUBackbone=Mistral-12B2026.04 | 21.2 | 4 | |
| TOFUBackbone=LLAMA-3.1-8B2026.04 | 20.9 | 4 | |
| GEMBackbone=Pythia-12B2026.04 | 17.6 | 3.5 | |
| BASEBackbone=PHI-4-14B2026.04 | 17.6 | 3.9 | |
| TOFUBackbone=Pythia-12B2026.04 | 16.1 | 3.5 | |
| BASEBackbone=OLMo-2-13B2026.04 | 13.9 | 3.8 | |
| BASEBackbone=Mistral-12B2026.04 | 12.7 | 3.8 | |
| BASEBackbone=LLAMA-3.1-8B2026.04 | 12.2 | 3.5 | |
| BASEBackbone=Pythia-12B2026.04 | 8.4 | 2.6 | |
| λ-PRBackbone=QWEN-3-8B2026.04 | 4.6 | 2.6 | |
| λ-PRBackbone=PHI-4-14B2026.04 | 3.3 | 2.4 | |
| λ-PRBackbone=OLMo-2-13B2026.04 | 3.2 | 2.3 | |
| λ-PRBackbone=Pythia-12B2026.04 | 3 | 2 | |
| λ-PRBackbone=Mistral-12B2026.04 | 2.9 | 2 | |
| λ-PRBackbone=LLAMA-3.1-8B2026.04 | 2.8 | 2.1 |