Receipts: paste text, get every claim marked by whether a source backs it up

说明与用途

Receipts 自动验证文本中的声明，输出 SUPPORTED、CONTRADICTED 或 UNSUPPORTED，并明确指出不确定性，帮助作者避免易被忽视的合理错误。

问题

Fast‑published text contains plausible errors that readers overlook, such as wrong dates or figures, and existing claim checkers often report “unsupported” as false, giving a false sense of confidence.

结果

Receipts automatically verifies claims in text, distinguishing supported, contradicted, and unsupported cases, and explicitly indicates uncertainty, thereby reducing the risk of plausible errors slipping through.

工作流

• Paste text into the browser demo or run the CLI
• The model extracts atomic, stand‑alone claims and proposes a Wikipedia search query
• The query retrieves top Wikipedia pages via the open API
• The tool filters retrieved sentences by vocabulary overlap with the claim
• The model judges the claim against the filtered passages, returning a verdict and the passage used
• The result is displayed with a clear label (SUPPORTED/CONTRADICTED/UNSUPPORTED) and source URL

证据

Receipts reads a block of text, pulls out the claims that a source could settle, looks each one up, and marks it one of three ways: SUPPORTED, CONTRADICTED, UNSUPPORTED

来源

The product is the third one. Most tools quietly treat "I could not find it" as "false," which means the tool is confident about things it never checked.

来源

The model compares the claim against those passages and returns a verdict plus the passage it used. It is told, explicitly, to use only the passages and never its own knowledge.

来源

The failure mode of a checker like this is a model "correcting" text from vibes. Every design choice is aimed at that: claims must stand alone, the judge sees only retrieved passages, the retrieved passages are filtered by vocabulary overlap with the claim, and a claim's numbers must match to the precision the claim states.

来源