Summary of In-context Learning Agents Are Asymmetric Belief Updaters, by Johannes A. Schubert et al.
In-context learning agents are asymmetric belief updaters
by Johannes A. Schubert, Akshay K. Jagadish, Marcel Binz, Eric Schulz
First submitted to arxiv on: 6 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 Large language models (LLMs) have unique learning dynamics when faced with instrumental learning tasks. Researchers found that LLMs update their beliefs asymmetrically, learning more from better-than-expected outcomes than worse ones. This effect reverses when learning about counterfactual feedback and disappears without agency implied. Idealized in-context agents derived through meta-reinforcement learning exhibited similar patterns. The study contributes to understanding in-context learning by highlighting the significant influence of problem framing on the learning process, a phenomenon also observed in human cognition. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models learn in interesting ways! When given tasks to complete, they update their ideas more when things go better than expected rather than worse. This changes when they’re told what could have happened instead. And if there’s no one “in charge” of the task, it all works differently again. Scientists made simple computer programs that acted like these language models and got similar results. |
Keywords
* Artificial intelligence * Reinforcement learning
