Hugging Face Agents Course · Final Project

GAIA Browser Research Agent

A transparent, serverless agent that loads the official evaluation set, applies deterministic tools first, optionally uses a Hugging Face model, and keeps every answer under human review.

Runs in your browserNo paid hardwareEditable answersOfficial scoring API
Ready. Load the official questions to begin.
Evaluation workspace 0 questions
#Task IDQuestionFileAnswer
No questions loaded yet.

The deterministic layer handles reversed instructions, operation-table symmetry checks, and strict botanical classification. Other questions can be sent to the optional inference model one at a time. Review the results before submission. Assignment guide · API docs