Long-context Understanding on Infini-Bench (test)
17.26Math ScoreSHAREDLLM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SHAREDLLMBase Model=LLaMA-32026.03 | 17.26 | 36.99 | 97.31 | |
| SHAREDLLMBase Model=Mistral-7B2026.03 | 16.12 | 29.8 | 65.73 | |
| Activation BeaconBase Model=LLaMA-32026.03 | 15.56 | 37.17 | 95.18 | |
| Activation BeaconBase Model=Mistral-7B2026.03 | 14.72 | 28.71 | 62.37 | |
| SHAREDLLMBase Model=LLaMA-22026.03 | 13.82 | 33.65 | 82.79 | |
| Activation BeaconBase Model=LLaMA-22026.03 | 12.14 | 32.05 | 80.58 | |
| LongAlpaca-16KBase Model=LLaMA-32026.03 | 12.03 | 25.28 | 16.13 | |
| StreamingLLMBase Model=LLaMA-32026.03 | 11.27 | 35.81 | 52.85 | |
| SnapKVBase Model=LLaMA-22026.03 | 9.95 | 28.83 | 2.31 | |
| BaseBase Model=LLaMA-32026.03 | 9.93 | 24.17 | 49.85 | |
| OmniKVBase Model=LLaMA-22026.03 | 8.81 | 26.25 | 3.66 | |
| OmniKVBase Model=LLaMA-32026.03 | 8.29 | 21.16 | 41.1 | |
| SnapKVBase Model=LLaMA-32026.03 | 7.82 | 17.73 | 43.51 | |
| StreamingLLMBase Model=Mistral-7B2026.03 | 7.26 | 18.84 | 9.75 | |
| LongAlpaca-16KBase Model=LLaMA-22026.03 | 6.23 | 25.74 | 4.87 | |
| StreamingLLMBase Model=LLaMA-22026.03 | 6 | 32.31 | 5.23 | |
| LongAlpaca-16KBase Model=Mistral-7B2026.03 | 5.41 | 21.19 | 12.48 | |
| SnapKVBase Model=Mistral-7B2026.03 | 4.73 | 16.18 | 15.71 | |
| OmniKVBase Model=Mistral-7B2026.03 | 3.81 | 19.77 | 14.98 | |
| BaseBase Model=Mistral-7B2026.03 | 3.57 | 20.65 | 5.41 | |
| BaseBase Model=LLaMA-22026.03 | 2.85 | 22.79 | 1.85 |