Summarization on Summarization (Rouge-L)
44.99Rouge-LTeacher (70B SFT)
Evaluation Results
| Method | Links | |
|---|---|---|
| Teacher (70B SFT)Model Size=70B2026.06 | 44.99 | |
| 2-StageModel Size=8B2026.06 | 44.29 | |
| 2-StageModel Size=3B2026.06 | 44.02 | |
| MixModel Size=8B2026.06 | 43.4 | |
| 2-StageModel Size=1B2026.06 | 42.84 | |
| GKDModel Size=8B2026.06 | 42.41 | |
| MixModel Size=3B2026.06 | 42.17 | |
| SFTModel Size=8B2026.06 | 41.57 | |
| GKDModel Size=3B2026.06 | 41.37 | |
| SFTModel Size=3B2026.06 | 40.98 | |
| MixModel Size=1B2026.06 | 40.48 | |
| GKDModel Size=1B2026.06 | 38.63 | |
| SFTModel Size=1B2026.06 | 37.03 | |
| Baichuan2-Turbo-192KContext Window Size=192K2023.11 | 18.4 | |
| ChatGLM3-6B-32KContext Window Size=32K, Backbone=6B2023.11 | 17.8 | |
| ChatGLM2-6B-32KContext Window Size=32K, Backbone=6B2023.11 | 16.1 | |
| GPT3.5-Turbo-16KContext Window Size=16K2023.11 | 16 | |
| Our modelContext Window Size=8K2023.11 | 15.6 | |
| Vicuna-v1.5-7B-16KContext Window Size=16K, Backbone=7B2023.11 | 15.1 | |
| Qwen-14B-ChatContext Window Size=8K, Backbone=14B2023.11 | 13.9 | |
| InternLM-7B-8KContext Window Size=8K, Backbone=7B2023.11 | 12.4 | |
| Longchat-v1.5-7B-32KContext Window Size=32K, Backbone=7B2023.11 | 9.9 | |
| Xgen-7B-8KContext Window Size=8K, Backbone=7B2023.11 | 2.2 |