forked from academicpages/academicpages.github.io
-
Notifications
You must be signed in to change notification settings - Fork 5
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
3 changed files
with
23 additions
and
2 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,21 @@ | ||
--- | ||
title: "Guideline Learning for In-Context Information Extraction" | ||
collection: publications | ||
permalink: /publication/2023-11-01-EMNLP | ||
excerpt: 'We propose to learn the task description (guideline) to learn from training samples for in-context learning' | ||
date: 2023-11-01 | ||
venue: 'EMNLP' | ||
paperurl: 'https://aclanthology.org/2023.emnlp-main.950/' | ||
citation: 'Chaoxu Pang, Yixuan Cao, Qiang Ding, and Ping Luo. 2023. Guideline Learning for In-Context Information Extraction. In EMNLP, 2023.' | ||
--- | ||
|
||
Machine learning algorithms typically learn by updating parameters. But in the era of large models, updating parameters becomes infeasible for many closed-source models. Therefore, how can learning be achieved without updating model parameters, thereby enhancing the model's in-context learning capabilities? | ||
|
||
We have proposed a **Guideline Learning** framework based on in-context learning. A guideline refers to text used to describe tasks, similar to the instructions given to annotators during the labeling process. This framework can generate and follow guidelines, enabling the model to learn a series of task guidelines from error samples without updating its parameters. Experiments in event extraction and relation extraction have shown that Guideline Learning improves the performance of in-context learning in information extraction tasks. This suggests that learning task instructions without updating parameters can serve as a new approach to learning. | ||
|
||
|
||
[The Guideline Learning Framework.](/images/guideline-learning.png) | ||
On the leftside of the figure, during inference, given an input query, the framework retrieve relevant guideline rules to conduct in-context learning. On the rightside, during training, given an input query, the prediction and the annotation are feed into the LLM to refect and generate (update) relavant guideline rules. The training output is a set of guideline rules. | ||
|
||
|
||
[paper](https://aclanthology.org/2023.emnlp-main.950/) |
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.