Open RL Write-up Released: Environment, Hand-Rated Pool, and All Paintings

Key Info

A full write-up is now available, sharing an open RL environment, a hand-rated evaluation pool, three training runs, and every painting generated during the project.

Highlights

• Uses TRL (Transformer Reinforcement Learning) with an open environment for the training setup.
• Open artifacts include code, models, dataset, RL environment, and all generated paintings.
• The hand-rated pool provides human-evaluation data across the documented runs.
• A follow-up blog with the full open artifacts is expected soon.