[Badcase]: 相同的微调数据，Qwen1.5 14B准确率比Qwen2.5 14B高20%左右，这是什么原因

[Jayc-Z]

Model Series

Qwen2.5

What are the models used?

qwen2.5-14B-Instruct

What is the scenario where the problem happened?

Qwen2.5-14B-Instruct LoRA微调效果不好

Is this badcase known and can it be solved using avaiable techniques?

• I have followed the GitHub README.
• I have checked the Qwen documentation and cannot find a solution there.
• I have checked the documentation of the related framework and cannot find useful information.
• I have searched the issues and there is not a similar one.

Information about environment

任务简介：判断query的最小时间单位，方便环比同比的时间推理
例：
query：我要查询本月的每天电力智能通信网关的环比pr值
分析：query中的最小时间单位为天，因此unit=day，一句话中需要查询多个时间，因此"is_multi": "multi"
response：{"unit": "day", "is_multi": "multi"}
实验描述：用同样的1000条数据集，超参设置learning_rate: 0.00005，num_train_epochs: 24，Qwen1.5-14B-Chat准确率达到96%，但Qwen2.5-14B-Instruct只有76%，想请教一下为什么会退化这么多，微调时有什么需要注意的点。

Description

Steps to reproduce

This happens to Qwen2.5-xB-Instruct-xxx and xxx.
The badcase can be reproduced with the following steps:

1. ...
2. ...

The following example input & output can be used:

system: ...
user: ...
...

Expected results

The results are expected to be ...

Attempts to fix

I have tried several ways to fix this, including:

1. adjusting the sampling parameters, but ...
2. prompt engineering, but ...

Anything else helpful for investigation

I find that this problem also happens to ...

[LaoLiulaoliu]

分别画一下loss曲线。