An offline 0.8B model that ties Gemini Flash on OpenStreetMap questions
Source
jonready.com
Author
mips_avatar
Date
Why it matters
Shows that SFT plus RL can lift a stock 0.8B model from 14 to 83 on a tool-using retrieval task, making on-device groundingTying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.Full definition → agents plausible.
Terms in this piece · Glossary
grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
