Recall Sloth: nobody reads recall notices, so Gemma 4 reads 3,411 of them in your browser
Описание и применение
Recall Sloth позволяет пользователям быстро проверить, попадают ли их покупки в списки рекламаций, используя Gemma 4 E2B в браузере.
Задача
Consumers rarely read U.S. consumer product recall notices, making it hard to know if purchased items are recalled.
Результат
Recall Sloth achieves 95% recall detection on 258 positive cases, with zero false dismissals and only 5% false alarms, outperforming rule‑only and other LLM setups.
Рабочий процесс
- Paste a list of purchased items, an Amazon "Your Orders" page, or order emails into the web interface.
- Gemma 4 E2B runs in a Web Worker using Transformers.js and ONNX Runtime Web on WebGPU.
- The tool matches each item to recall candidates by brand and product type.
- If the recall’s model number is present in the item, the item is marked "Recalled."
- If the brand and type match but the model number differs, the item is marked "Check the label" with the notice’s guidance.
- If the brand matches but the product type differs, the item is marked "Not this one."
- If no recall shares the brand and type, the item is marked "No recall on file."
- Results are displayed instantly; nothing leaves the browser tab.
- Users can review the four piles and take appropriate action.
- The page is static, served from Cloudflare; recall data is fetched from the CPSC public API to stay current without a server.
Доказательства
Gemma 4 E2B: is it even the same kind of product as the recalled one?
The page is static files on Cloudflare's free tier. Gemma 4 E2B runs in a Web Worker through Transformers.js and ONNX Runtime Web on WebGPU.
Gemma 4 E2B took false alarms from 16 to 5 without dismissing a single recalled product, and the bigger E4B was no better.
The recall list ships with the page, and the page also asks CPSC's public API directly for anything published since, so it stays current without a server or a cron job.
- Дата записи: 2026-10-21T19:11:38.415Z