Frontend
Clean Tickets, Wrong Orders: Catching Silent Errors with a Language Model in Your Browser Tab
arXiv cs.HC0 views
arXiv:2611.18284v1 Announce Type: cross Abstract: A small language model that turns a typed order into structure can fail in two ways. It can print something broken, which any system rejects. Or it can print a clean, well formed order that nobody placed. Our companion work calls the second kind a silent error, and it is the dangerous one. We present a web demo in which a fine-tuned 0.8B parameter model runs entirely inside the visitor's browser tab, through llama.cpp compiled to WebAssembly. No order text leaves the device. The page parses orders as the visitor types, scores every ticket with the PIZZA benchmark's own matcher ported to TypeScript, refuses anything off the menu with "?" so the model can face the open web, and lets visitors hunt for silent errors or try to cause one. Run in the browser on all 1,357 PIZZA test orders, the deployed model writes the same output as the server evaluation for 99.0% of them and reaches the same exact match (0.882); we also report its latency on 3 consumer laptops. Demo: https://wrong.cuyor.com
Source:https://arxiv.org/abs/2611.18284PreviousDECODE: Diagnosing Interpretability Failures in Automated Interpretation ProcessNextToward On-Device Markerless 6DoF Tracking of a Physical Manipulative on a Mixed-Reality Headset: A Preliminary Study
Related posts
OOPs in Java
0dev.to Java
Taking over a codebase you don't know: where to start
0dev.to Git
Automating Price Validation Across 633 Static Pages Before Every Deploy
0dev.to TypeScript
How We Use Flight Data APIs to Track Every Airport Pickup
0dev.to TypeScript
How to export a Framer site to TanStack Start
0dev.to TypeScript
LUNAR BROWSER :
0dev.to TypeScript