2025-11-05 –, Eng
Let's get back to the basic with an energized 5 minutes talk about what tool use in LLMs really is, how it started, how it's measured, the Gorilla paper that started it all, and the difference between function calling and strucutured output.
We'll cover the basic idea - how we lived before the LLM vendors let us expclicitly pass the signature of the function (we did prompt engineering) ; how it's implemented under the hood (it's prompt engineering all the way down) ; what should someone know when approaching this (use function calling it's the best way to force an LLM to do something, better than threatning to shoot an intern if the LLM fails to provide the required schema) ; if we have time we'll dive shallow into the world of evaluating tool use, which is useful also for evaluating the most hype-y thing of all - Agents and MCP.
No previous knowledge expected
Over a decade of industry experience in the Intersection of DS/ML and Cyber; in the last year I'm working on LLM Apps @ Blinkops .