doocs / doocs/cose

[Bug]: 今日头条latex公式丢失,无法解析

Open
#254 0 comments 0 reactions 0 assignees View on GitHub
bug
Dominant language
JavaScript
Stars
744
Forks
125
PR merge metrics
No merged PRs in 30d

Description

### Description / 问题描述

今日头条对于latex公式完全不支持,全部无法解析,丢失了

### Steps to Reproduce / 复现步骤
```
如上图所示,整个 SUAN 架构分为输入层、多层堆叠的统一注意力模块(UAB),以及负责汇总打分的最终预测层。在预测层中,模型提取了最后一个UAB的输出端 candidate 表征,与展平后的用户静态画像(User Profile)和其他细粒度辅助特征一并送入全连接网络(MLP):

$$
\begin{equation}
\hat{y} = \sigma(z), \quad z = \mathrm{MLP}(E_{\mathrm{block}}[-1,:], e_p, e_{\mathrm{other}}),
\end{equation}
$$

其中 $\sigma$ 为 sigmoid 激活概率分布函数,$z$ 是输出的 logits 评分。

在这套计算层中,真正的魔法引擎在于中间的那一摞堆叠起来的 **统一注意力模块 (UAB)**。
```

### Expected Behavior / 预期行为

Image

### Suggested Implementation / 实现建议

_No response_

### Platform / 平台

头条

### Browser / 浏览器

Chrome

### Extension Version / 扩展版本

1.3.6

### Environment Details / 环境信息

macOS 15.7.7
谷歌浏览器 149.0.7827.197

### Screenshots / 截图

_No response_

### Additional Context / 补充信息

_No response_

Contributor guide

Open the contributing guide

Research direction

Reproduce the issue in Chrome on Toutiao using the supplied SUAN architecture text and LaTeX snippet with extension version 1.3.6. Compare the published result with the expected behavior shown in the issue and identify the extension entry point responsible for handling LaTeX before deciding what successful parsing should look like.

Written by the indexing model from the issue text.

Assessment

Tech stack
javascript, latex
Domain
frontend
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.