Downstream Task on HellaSwag
76.7AccuracyPhi-3 Mini-4k
Evaluation Results
| Method | Links | |
|---|---|---|
| Phi-3 Mini-4kShots=5-shot2024.10 | 76.7 | |
| Llama 3 8BParameters=8B, Shots=5-shot2024.10 | 71.1 | |
| Mistral 8x7BParameters=8x7B, Shots=5-shot2024.10 | 70.4 | |
| Mistral 7BParameters=7B, Shots=5-shot2024.10 | 58.5 | |
| Shakti-LLMParameters=2.5B, Shots=5-shot2024.10 | 52.4 | |
| Gemma 7BParameters=7B, Shots=5-shot2024.10 | 49.8 | |
| TinyLlama-1.1BFFN Size=100%, Shots=102025.06 | 46.45 | |
| DIVE 1/8FFN Size=50% × 1, Shots=102025.06 | 37.09 | |
| FLAPFFN Size=50%, Shots=102025.06 | 35.37 | |
| LLaMA-MoE 1/8FFN Size=50% × 1, Shots=102025.06 | 33.24 | |
| LLM-PrunerFFN Size=50%, Shots=102025.06 | 33.17 | |
| DIVE 2/8FFN Size=25% × 2, Shots=102025.06 | 32.86 | |
| LLaMA-MoE 2/8FFN Size=25% × 2, Shots=102025.06 | 31.93 |