Multi-document Question Answering on Chinese Multi-doc QA (test)
44.8Rouge-LChatGLM3-6B-32K
Evaluation Results
| Method | Links | |
|---|---|---|
| ChatGLM3-6B-32KContext Window Size=32K, Backbone=6B2023.11 | 44.8 | |
| Our modelContext Window Size=8K2023.11 | 44.6 | |
| ChatGLM2-6B-32KContext Window Size=32K, Backbone=6B2023.11 | 37.6 | |
| Baichuan2-Turbo-192KContext Window Size=192K2023.11 | 36.8 | |
| GPT3.5-Turbo-16KContext Window Size=16K2023.11 | 28.7 | |
| Longchat-v1.5-7B-32KContext Window Size=32K, Backbone=7B2023.11 | 19.5 | |
| Vicuna-v1.5-7B-16KContext Window Size=16K, Backbone=7B2023.11 | 19.3 | |
| Qwen-14B-ChatContext Window Size=8K, Backbone=14B2023.11 | 18.7 | |
| InternLM-7B-8KContext Window Size=8K, Backbone=7B2023.11 | 16.3 | |
| Xgen-7B-8KContext Window Size=8K, Backbone=7B2023.11 | 11 |