Summary of Are Language Models Rational? the Case Of Coherence Norms and Belief Revision, by Thomas Hofweber et al.
Are language models rational? The case of coherence norms and belief revision
by Thomas Hofweber, Peter Hase, Elias Stengel-Eskin, Mohit Bansal
First submitted to arxiv on: 5 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- 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 investigates whether norms of rationality apply to machine learning models, specifically language models. The researchers focus on coherence norms, which measure the logical consistency or strength of belief within a model. They introduce the Minimal Assent Connection (MAC) and propose a new account of credence, assigning strength of belief based on internal next-token probabilities. The study finds that rational norms tied to coherence apply to some language models but not others, highlighting implications for AI safety, alignment, and understanding model behavior. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper asks if machine learning models, like language models, follow rules of reason and logic. Researchers looked at how well these models make sense and are consistent with themselves. They came up with a new way to measure how strongly a model believes something is true based on the probabilities it gives for each possible next step. The study shows that some language models do follow logical rules, but others don’t. This matters because understanding how AI models think can help us make them safer and more trustworthy. |
Keywords
* Artificial intelligence * Alignment * Machine learning * Token
