Commit
•
da91ceb
1
Parent(s):
f201cac
Update README.md
Browse files
README.md
CHANGED
@@ -10,9 +10,31 @@ pretty_name: bridge
|
|
10 |
size_categories:
|
11 |
- n<1K
|
12 |
---
|
|
|
|
|
13 |
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
-
The dataset targets scenarios where the student makes a math mistake.
|
17 |
|
18 |
For more information about how the dataset is curated, please check out our codebase: https://github.com/rosewang2008/bridge/, and paper: https://arxiv.org/pdf/2310.10648
|
|
|
10 |
size_categories:
|
11 |
- n<1K
|
12 |
---
|
13 |
+
TLDR: This dataset is a real-world math tutoring dataset from the NAACL 2024 paper ``Bridging the Novice-Expert Gap via Models of Decision-Making: A Case Study on Remediating Math Mistakes''.
|
14 |
+
The dataset targets scenarios where the student makes a math mistake.
|
15 |
|
16 |
+
# 🌁 Bridging the Novice-Expert Gap via Models of Decision-Making
|
17 |
+
|
18 |
+
[Paper Link](https://arxiv.org/abs/2310.10648), [Code Link](https://github.com/rosewang2008/bridge/)
|
19 |
+
|
20 |
+
**NAACL 2024**
|
21 |
+
|
22 |
+
**Title:** Bridging the Novice-Expert Gap via Models of Decision-Making: A Case Study on Remediating Math Mistakes
|
23 |
+
|
24 |
+
**Authors:** Rose E. Wang, Qingyang Zhang, Carly Robinson, Susanna Loeb, Dorottya Demszky
|
25 |
+
|
26 |
+
**Main Idea: We contribute Bridge 🌁, a method that uses cognitive task analysis to translate an expert's implicit thought process into an explicit decision-making model**.
|
27 |
+
|
28 |
+
Scaling high-quality tutoring remains a major challenge in education.
|
29 |
+
Due to growing demand, many platforms employ novice tutors who, unlike experienced educators, struggle to address student mistakes and thus fail to seize prime learning opportunities.
|
30 |
+
Our work explores the potential of large language models (LLMs) to close the novice-expert knowledge gap in remediating math mistakes.
|
31 |
+
**Bridge 🌁 leverages cognitive task analysis to model an expert's internal decision-making in remediation: Experts internally identify (A) the student's error, (B) a remediation strategy, and (C) their intention before generating a response.**
|
32 |
+
We construct a dataset of 700 real tutoring conversations, annotated by experts with their decisions.
|
33 |
+
We evaluate state-of-the-art LLMs on our dataset and find that the expert's decision-making model is critical for LLMs to close the gap:
|
34 |
+
responses from GPT4 with expert decisions (e.g., ``simplify the problem'') are +76% more preferred than without.
|
35 |
+
Additionally, context-sensitive decisions are critical to closing pedagogical gaps:
|
36 |
+
random decisions decrease GPT4's response quality by -97% than expert decisions.
|
37 |
+
Our work shows the potential of embedding expert thought processes in LLM generations to enhance their capability to bridge novice-expert knowledge gaps.
|
38 |
|
|
|
39 |
|
40 |
For more information about how the dataset is curated, please check out our codebase: https://github.com/rosewang2008/bridge/, and paper: https://arxiv.org/pdf/2310.10648
|