Datasets:
Add Github repo link and data details
Browse filesThis PR adds a link to the associated Github repository and details about the datasets provided.
README.md
CHANGED
|
@@ -3,12 +3,19 @@ license: mit
|
|
| 3 |
task_categories:
|
| 4 |
- question-answering
|
| 5 |
---
|
|
|
|
| 6 |
This dataset accompanies the paper:
|
| 7 |
|
| 8 |
**[When Do LLMs Admit Their Mistakes? Understanding the Role of Model Belief in Retraction](https://arxiv.org/abs/2505.16170)**
|
| 9 |
|
| 10 |
It includes the original Wikidata questions used in our experiments, with train/test split. For a detailed explanation of the dataset construction and usage, please refer to the paper.
|
| 11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
## Citation
|
| 13 |
```bibtex
|
| 14 |
@misc{yang2025llmsadmitmistakesunderstanding,
|
|
|
|
| 3 |
task_categories:
|
| 4 |
- question-answering
|
| 5 |
---
|
| 6 |
+
|
| 7 |
This dataset accompanies the paper:
|
| 8 |
|
| 9 |
**[When Do LLMs Admit Their Mistakes? Understanding the Role of Model Belief in Retraction](https://arxiv.org/abs/2505.16170)**
|
| 10 |
|
| 11 |
It includes the original Wikidata questions used in our experiments, with train/test split. For a detailed explanation of the dataset construction and usage, please refer to the paper.
|
| 12 |
|
| 13 |
+
This repository provides:
|
| 14 |
+
* **Original datasets:** `wikidata_{train,test}_free.jsonl`, `celebrity_{train,test}_free.jsonl`.
|
| 15 |
+
* **Continuation datasets:** for Llama3.1-8B, Qwen2.5-7B, and Olmo2-7B.
|
| 16 |
+
|
| 17 |
+
Code: https://github.com/YuqingYang125/retraction
|
| 18 |
+
|
| 19 |
## Citation
|
| 20 |
```bibtex
|
| 21 |
@misc{yang2025llmsadmitmistakesunderstanding,
|