Summary of Investigating Recurrent Transformers with Dynamic Halt, by Jishnu Ray Chowdhury et al.
Investigating Recurrent Transformers with Dynamic Halt
by Jishnu Ray Chowdhury, Cornelia Caragea
First submitted to arxiv on: 1 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
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 the inductive biases of two approaches to augmenting Transformers with recurrent mechanisms. The first approach incorporates a depth-wise recurrence similar to Universal Transformers, while the second uses a chunk-wise temporal recurrence like Temporal Latent Bottleneck. Novel combinations and extensions are also proposed and evaluated. Diagnostic tasks such as Long Range Arena (LRA), flip-flop language modeling, ListOps, and Logical Inference are used to probe the inductive biases of the models. The code is released for public use. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how two ways to add recurrence to Transformers work together. They’re trying to figure out if one way (like Universal Transformers) or another way (like Temporal Latent Bottleneck) is better, and what happens when you mix them together. They test these models on some special tasks that help us understand their strengths and weaknesses. |
Keywords
* Artificial intelligence * Inference
