Summary of Clarify: Improving Model Robustness with Natural Language Corrections, by Yoonho Lee et al.
Clarify: Improving Model Robustness With Natural Language Corrections
by Yoonho Lee, Michelle S. Lam, Helena Vasconcelos, Michael S. Bernstein, Chelsea Finn
First submitted to arxiv on: 6 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
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 In this paper, researchers propose Clarify, a new approach to teaching machine learning models that leverages human feedback at the concept level. Unlike previous methods that require manual labeling of training data, Clarify allows users to provide short text descriptions of model misconceptions, which are then used to improve the training process. The authors demonstrate the effectiveness of Clarify in two datasets and conduct a case study on ImageNet, finding and rectifying 31 novel hard subpopulations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning models can learn incorrect ideas if they’re taught with misleading data. To fix this, we need to give them extra information. One way is by adding labels for things that are wrong or providing corrected data. But this takes a lot of work. The researchers think people can do better than that. They propose an easy way to correct model mistakes using text descriptions. It’s called Clarify. Users just write a short sentence about what the model gets wrong, and then the computer fixes it automatically. This is the first system that lets users correct models in one step. Studies show that regular people can use Clarify to make models better. |
Keywords
* Artificial intelligence * Machine learning
