Spaces:
Running
Babbling Model
Link: https://hf.co/chat/r/5gOCSbi
The model seems to be babbling, and is always long-winded when responding to simple sentences, which is nonsense. (Just output one core sentence, the rest is either off-topic or gibberish)
As the quoted example shows, I actually asked a question, the rest is in the evaluation of the answer.
For example, Do you know the game Arknights's character Amiya? The answer to this question is off-topic. (Actually, some models also address this question, not a fault unique to Huggingchat.) The correct answer is: Amiya, leader of Rhodes Island ...... (The name of the game is correct, there is no need to read the introduction to this)
The following is the original Chinese text, translated above via DeepL.
这个模型似乎是在胡言乱语,在回复简单句时总是长篇大论,就是废话。(输出一句最核心的就行了,其他内容不是离题万里就是胡言乱语)
如引用的例子来看,其实我就问了一个问题,其他都是在对回答的评价。
例如,Do you know the game Arknights's character Amiya?这个问题的答案是离题万里的。(其实一些模型也涉及此问题,不是Huggingchat独有毛病。)正确答案是:Amiya, leader of Rhodes Island......(游戏名是正确的,对此介绍不用看了)
We did some adjustments, feel free to retry and do feedback 👍