Summary of Assessing Episodic Memory in Llms with Sequence Order Recall Tasks, by Mathis Pink et al.
Assessing Episodic Memory in LLMs with Sequence Order Recall Tasks
by Mathis Pink, Vy A. Vo, Qinyuan Wu, Jianing Mu, Javier S. Turek, Uri Hasson, Kenneth A. Norman, Sebastian Michelmann, Alexander Huth, Mariya Toneva
First submitted to arxiv on: 10 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 introduces Sequence Order Recall Tasks (SORT) as a new evaluation framework for evaluating long-term episodic memories in Large Language Models (LLMs). Existing benchmarks primarily focus on assessing semantic aspects of memory, whereas SORT requires LLMs to recall the correct order of text segments. The authors adapt tasks from cognitive psychology to study episodic memory and demonstrate that humans can recall sequence orders based on long-term memory of a book. They also show that while models perform well when presented with in-context text during evaluation, their performance drops significantly when trained only on the book text. SORT aims to aid in the development of memory-augmented models by evaluating more aspects of memory. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to test how good Large Language Models (LLMs) are at remembering things that happened before. Right now, tests mostly focus on how well they understand what something means, not how they remember when and where it happened. The authors take ideas from studying human memory and create a new kind of test that asks LLMs to put text in the correct order. They show that humans can do this by remembering a book. The models are good at it when given extra context during the test, but not so good when only trained on the book text. This new way of testing might help make LLMs better at remembering things. |
Keywords
» Artificial intelligence » Recall
