bytedance / bytedance/GUI-ReWalk
关于 Table 2 中 AndroidControl-Low 指标(Type Acc. 与 SR)的疑问
- Dominant language
- Python
- Stars
- 41
- Forks
- 7
- PR merge metrics
- No merged PRs in 30d
Description
您好,感谢作者开源论文/项目。
我对 Table 2 中 GUI-ReWalk-7B 在 AndroidControl-Low 上的结果有些疑问:
- Type Acc. = 91.7
- SR = 96.3
这里我可能理解错了指标定义,但我有些困惑:为什么 SR 会高于 Type Acc.?
想请教一下:
1. “Type Acc.” 的具体定义是什么?
2. “SR” 的具体定义是什么?
3. 4.2 节里提到的 “step success rate” 是否就是表 2 里的 “SR”?
4. 在什么情况下会出现 SR 高于 Type Acc. 的结果?
如果这是正常现象,是否可以补充一下说明?谢谢!
Contributor guide
Research direction
Start with the definitions in Table 2 and the discussion of “step success rate” in Section 4.2. Compare Type Acc. and SR as described there, then clarify whether they measure the same unit and when SR can exceed Type Acc. Done means the metric definitions and their relationship are explained consistently in the project documentation.
Written by the indexing model from the issue text.
Assessment
- Domain
- documentation
- Issue type
- Documentation
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100