2023 LLM model Translation Survery Results
Results from my 2023 Survey on Translation using different LLM models
In 2023 I made a survey asking netizens to rate the translation of a paragraph of an XianXia novel by different LLM models to see how they fare against the Human translation.
In total about 30 readers conducted the survey (thanks a lot !) and here are the results:
I'll probably do another one for models in 2024 with the same prompt and text, to see how much it has evolved.
survey link: https://form.jotform.com/233077764011048
|
Model Name |
Approval Rating |
|
Falcon-180B |
65.52% |
|
Vicuna-33B |
65.52% |
|
llama-2-70b-chat |
69.66% |
|
GPT4 2023 |
58.62% |
|
Qwen-14B |
52.41% |
|
Human Translation |
79.31% |
|
Claude2 2023 |
46.21% |
|
Google Translate |
66.90% |
|
Bard 2023 |
78.62% |