Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,15 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
This classifier is trained on LLM responses designed to capture a model refuting a false claim.
|
| 8 |
+
|
| 9 |
+
They're prompt results of OpenAI gpt-3.5-turbo run on June 1 2023.
|
| 10 |
+
Prompts are constructed by prepending "Explain why" to claims made in the True-False dataset provided alongside [The Internal State of an LLM Knows When its Lying](https://arxiv.org/abs/2304.13734).
|
| 11 |
+
Only the first sentence of the response is included (from nltk.sent_tokenize).
|
| 12 |
+
The original labels are used, where 0 corresponds to a false claim.
|
| 13 |
+
That is, the 0 labels should be almost all refutations.
|
| 14 |
+
Spurious and missing refutations were removed by hand, corresponding to about 10% of the data.
|
| 15 |
+
This were caused by either an incorrect model assertion, or errors in the source data.
|