v0.10.145

© ARISE Network

ARISEKit
NOHARMKit

Tutorial draft

Dashboard Quick Reference Framework

Pipeline

DatasetAdd your cases
RubricsBuild your rubrics
ResponsesTest your models
AutogradingReview judgments
ScoresCalculate scores
LeaderboardCompare models

1

Dataset

Build a benchmark from your data. Ensure protected health information (PHI) is removed, use a secure environment, or use public/synthetic data.

Task

The case, prompt, or question you will give to the language model (LLM).

Solution

Appropriate responses. "Add alternative solution" if multiple acceptable alternatives

Solution 1

Errors (optional but recommended)

Possible errors and their potential severity

Import dataset

Upload a CSV or JSON file with multiple cases. Your file should have 3 fields: Task, Solution, and Errors. For CSV, concatenate multiple acceptable solutions into one Solution field, with "SOLUTION:" as a separator. This file will be imported into your project for your coding agent to process

Project

Drop a CSV or JSON file here

NOHARM Rubric Editor

Now return to our pipeline.