Google Research· ResearchResearch paper

A collaborative approach to image generation

Google Research introduces PASTA, a reinforcement learning agent that collaborates with users over multiple turns to refine text-to-image outputs by learning user preferences, enabled by a novel user simulation technique. They also release a foundational dataset of over 7,000 human rater interactions.

From the source

We introduce PASTA, a reinforcement learning agent that refines text-to-image output over multiple turns of interaction with a user by learning their unique preferences. This process is made possible by a novel user simulation technique. Through human evaluations, we created a novel dataset of sequential preferences, which we then used to compare PASTA with a baseline state-of-the-art model. The results demonstrated that PASTA, trained with our mix of real and simulated data, consistently produced images that users rated as more satisfying. We’ve also released our foundational dataset with a collection of over 7,000 human rater interactions with PASTA.

research.google