Tutorials¶
Step-by-step walkthroughs with screenshots of Arena. Each tutorial shows where to click and what you should see on screen.
On Home, Get started has two cards: Setting up an experiment and Custom environment. Click either card to start that tour. These written tutorials cover the same flows with screenshots.
Area |
Start here |
|---|---|
Environments |
|
Datasets |
|
Experiments |
|
Agents |
|
Account |
LLM data¶
Tutorial |
What you will do |
|---|---|
Create a Language Reasoning dataset (Question / Answer) |
|
Create a Language Preference dataset (Prompt / Chosen / Rejected) |
|
Create a Language SFT dataset (Prompt / Target) |
|
Add a language-based gym for LLM simulation |
|
Create a non-tabular object detection dataset with mask targets, then preprocess it (Enterprise) |
LLM training¶
Tutorial |
What you will do |
|---|---|
Create a project with Advanced Training enabled and open Datasets |
|
SFT dataset + algorithm SFT through the wizard |
|
Preference dataset + DPO through the wizard and start training |
|
Reasoning dataset + CISPO, including prompt and reward validation |
|
Language-based gym + LLMPPO with |
|
Open metrics and checkpoint rows after a run |
Agents¶
Tutorial |
What you will do |
|---|---|
Deploy from a checkpoint and send a test prompt (UI, CLI, Python) |
Classic RL¶
Tutorial |
What you will do |
|---|---|
Create a Classic RL project, run the experiment wizard, train, open results |
|
Add custom gym code, set an entrypoint, validate |
Account¶
Tutorial |
What you will do |
|---|---|
Read usage and the balance (org Admin) |
Before you start¶
Sign in to Arena with an account that can create projects in your organization.
Training runs spend balance. The screenshots use sample org Acme Lab with stable demo data.