Summary of One-shot World Models Using a Transformer Trained on a Synthetic Prior, by Fabio Ferreira and Moreno Schlageter and Raghu Rajan and Andre Biedenkapp and Frank Hutter
One-shot World Models Using a Transformer Trained on a Synthetic Prior
by Fabio Ferreira, Moreno Schlageter, Raghu Rajan, Andre Biedenkapp, Frank Hutter
First submitted to arxiv on: 21 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes One-Shot World Model (OSWM), a transformer-based model that learns to represent real-world environments from purely synthetic data. OSWM is trained using Prior-Fitted Networks, which involves masking next-state and reward information at random context positions and querying the model to make predictions based on the remaining transition context. The model can quickly adapt to simple grid worlds and other environments by providing 1k transition steps as context, allowing it to train environment-solving agent policies. However, transferring OSWM to more complex environments remains a challenge. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about creating a computer model that can learn from pretend data, rather than real data. The goal is to make the model understand how different environments work, so it can solve problems in those environments. The model is called One-Shot World Model and it’s trained using special computer algorithms. It works by looking at some example transitions between states in an environment, and then trying to predict what would happen next based on that information. The model is pretty good at solving simple problems, but it gets stuck when faced with more complex environments. |
Keywords
* Artificial intelligence * One shot * Synthetic data * Transformer
