Combat LLM Spam by Building a Web of Trust

通过构建信任网络打击 LLM 垃圾内容

Tangled now has native support for vouching! You can vouch or denounce users that you interact with. Vouched users will have a green shield icon beside their profile pictures, and denounced users will have a red one. You can use this to inform decisions about an interaction. You can also see the vouch/denounce decisions made by your circle. Tangled 现在原生支持“担保”（vouching）功能！你可以为你互动的用户进行担保或谴责。获得担保的用户头像旁会显示一个绿色盾牌图标，而受到谴责的用户则会显示一个红色图标。你可以利用这些信息来辅助决策。此外，你还可以查看你信任圈内成员所做出的担保或谴责决定。

Why vouch?

为什么要进行担保？

Vouching serves as a signal of trust to your circle. The bar to submit code to a project has never been lower thanks to LLM based tooling. LLM tools are really good at creating “uncanny valley” submissions. Code that looks correct but is subtly wrong. The onus is on maintainers to now take the time to review such submissions. To ease this burden, maintainers from across the Tangled network can now vouch for or denounce contributors that misuse these tools and create a maintenance burden. 担保是向你的信任圈传递的一种信任信号。得益于基于 LLM 的工具，向项目提交代码的门槛从未如此之低。LLM 工具非常擅长制造“恐怖谷”式的提交——代码看起来正确，但实际上存在细微错误。现在，维护者必须花费大量时间来审查这些提交。为了减轻这一负担，Tangled 网络中的维护者现在可以对滥用这些工具并造成维护压力的贡献者进行担保或谴责。

Mindful design

深思熟虑的设计

Such systems need careful consideration. Vouching on Tangled includes the following to begin with: vouching/denouncing with a text-based reason field; attenuation: you can only view decisions made by you and your circle; no consequences to being denounced: at present, denounced users aren’t blocked from the project, but simply have a red warning label in parts of the UI. 此类系统需要谨慎考量。Tangled 上的担保功能初步包含以下特性：支持带有文本理由的担保/谴责；衰减机制：你只能查看你本人及其信任圈所做出的决定；被谴责者无后果：目前，被谴责的用户不会被项目封禁，仅会在 UI 的部分位置显示红色警告标签。

Some additions that I want to put in down the line: decay of vouches: maintainers and contributors tend to move on from projects over time, so vouches should decay as time passes, and be renewed every now and then; evidence trails: vouching for a user right after merging a PR should add the PR to the vouch record as a piece of evidence. 未来我计划加入的一些功能：担保衰减：维护者和贡献者往往会随时间推移而离开项目，因此担保应随时间推移而衰减，并需要不时更新；证据追踪：在合并 PR 后立即为用户担保，应将该 PR 作为证据添加到担保记录中。

How it works

工作原理

When you vouch for or denounce somebody on Tangled, you create a public record on your PDS. The record includes: whether you vouched for or denounced somebody; an optional reason for doing so. The Tangled appview then aggregates vouch data from across the network, and displays vouch “hats” over profiles at points of interaction: in issues and issue comments; in pull-requests and pull-request comments. 当你在 Tangled 上担保或谴责某人时，你会在你的 PDS（个人数据存储）上创建一个公开记录。该记录包括：你担保或谴责某人的状态，以及可选的理由。随后，Tangled 的 appview 会汇总全网的担保数据，并在交互点（如 Issue 及评论、Pull Request 及评论）的用户头像上显示担保“帽子”。

A hat appears over a user only if you have directly vouched/denounced them, or if somebody you have vouched for has vouched/denounced them. Additionally, there are no consequences for a denounced user. Only a hat. You can click on the hat to see who vouched/denounced this user in your circle. The consequences may change eventually, but for now you can use the hat to inform a decision. Start building your web of trust on Tangled today. 只有当你直接担保/谴责了某人，或者你担保过的人担保/谴责了某人时，该用户头像上才会出现“帽子”。此外，被谴责的用户不会受到任何惩罚，仅仅是显示一个图标。你可以点击该图标查看你信任圈内是谁对该用户进行了担保或谴责。虽然未来可能会引入惩罚机制，但目前你可以利用这个图标来辅助决策。今天就开始在 Tangled 上构建你的信任网络吧。