Latent Multi-hop Reasoning on 2WikiMultiHopQA
96.86PrecisionLlama 3.2-3B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama 3.2-3BPacking Strategy=Pack 62025.12 | 96.86 | 4.62 | 79.23 | |
| Gemma 2-2BPacking Strategy=Pack 62025.12 | 96.83 | 1.78 | 76.86 | |
| Llama 3.2-3BPacking Strategy=Pack 42025.12 | 96.66 | 4.55 | 78.53 | |
| Gemma 2-2BPacking Strategy=Pack 42025.12 | 96.53 | 1.32 | 76.68 | |
| Llama 3.2-3BPacking Strategy=Pack 2-42025.12 | 96.48 | 5.45 | 76.63 | |
| Llama 3.1-8BPacking Strategy=Pack 22025.12 | 96.42 | 21.9 | 77.83 | |
| Gemma 2-2BPacking Strategy=Pack 2-42025.12 | 96.37 | 2.36 | 74.85 | |
| Llama 3.2-3BPacking Strategy=Pack 4-82025.12 | 96.27 | 6.59 | 77.69 | |
| Llama 3.2-3BPacking Strategy=Pack 22025.12 | 96.26 | 5.24 | 76.35 | |
| Gemma 2-2BPacking Strategy=Pack 4-82025.12 | 96.23 | 2.26 | 75.42 | |
| Gemma 2-2BPacking Strategy=Pack 22025.12 | 96.18 | 2.03 | 75.06 | |
| Gemma 2-2BPacking Strategy=Pack 2-4-82025.12 | 96.15 | 2.26 | 75.54 | |
| Gemma 2-2BPacking Strategy=Pack 82025.12 | 96.13 | 2.77 | 75.63 | |
| Llama 3.2-3BPacking Strategy=Pack 2-4-82025.12 | 96.13 | 6.75 | 77.59 | |
| Gemma 2-2BPacking Strategy=Pack 102025.12 | 96.12 | 4.9 | 76.15 | |
| Llama 3.2-3BPacking Strategy=Pack 102025.12 | 96.08 | 8.46 | 77.33 | |
| Llama 3.2-3BPacking Strategy=Pack 82025.12 | 96.01 | 7.69 | 77.47 | |
| Gemma 2-2BPacking Strategy=Pack 2-82025.12 | 96 | 3.3 | 75.39 | |
| Llama 3.2-3BPacking Strategy=Pack 2-82025.12 | 95.99 | 8.69 | 75.99 | |
| Llama 3.2-3BPacking Strategy=No packing2025.12 | 94.37 | 21.83 | 62.32 | |
| Gemma 2-2BPacking Strategy=No packing2025.12 | 94.23 | 9.15 | 69.84 | |
| Llama 3.1-8BPacking Strategy=No packing2025.12 | 92.29 | 51.31 | 66.52 |