Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

I Built an AI That Checks Whether Digital Products Deliver What They Promise

Description and use

PromiseDiff automatically verifies that digital products meet listing promises, classifying each claim and providing evidence in a Markdown report before publication. It combines LLM semantic checks with deterministic PDF metadata extraction.

Problem

Digital‑product listings often promise specific content (e.g., page counts, file formats, worksheets) that the delivered ZIP or PDF does not match, leading to refunds and negative reviews.

Outcome

In a demo with a 48‑page A4 Budget Planner PDF, PromiseDiff confirmed the page count, flagged the missing US Letter edition, and exported a Markdown report. A benchmark of ten synthetic PDFs achieved 90% exact‑label accuracy, 100% contradiction precision, and 75% supported recall.

Workflow

  • Paste listing text into the UI
  • Upload the PDF or ZIP of the product
  • PromiseDiff parses claims with Gemma
  • Python validates measurable claims (page count, file existence, page size)
  • Gemma evaluates semantic claims against extracted text
  • Evidence validation checks filename, page, and exact quote
  • Results are displayed with evidence and exported as Markdown

Evidence

PromiseDiff shows what the files can back up, what they contradict, and what they simply cannot prove.

Source

missing evidence != contradiction

Source

exact-label accuracy = correct labels / total cases = 9 / 10 = 90%

Source

  • Record date: 2026-10-13T21:55:54.161Z