poppingtonic
commited on
Commit
•
05527b5
1
Parent(s):
a61707e
Update README.md
Browse files
README.md
CHANGED
@@ -3,9 +3,6 @@ license: mit
|
|
3 |
---
|
4 |
# Dataset Release: Indirect Object Identification
|
5 |
`mecha_ioi` is a pair of datasets tailored for the Indirect Object Identification task, where sentences are generated from the following set of templates:
|
6 |
-
The purpose of this dataset is to facilitate interpretability research, inspired by the paper
|
7 |
-
_Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small_, from Redwood Research. We are not affiliated with Redwood Research,
|
8 |
-
and release this dataset to contribute to the collective research effort behind understanding how Transformer language models perform this task.
|
9 |
|
10 |
- BABA
|
11 |
```
|
@@ -48,6 +45,9 @@ abba_templates = [
|
|
48 |
"Friends {A} and {B} found a {OBJECT} at the {PLACE}. {B} gave it to {A}",
|
49 |
]
|
50 |
```
|
|
|
|
|
|
|
51 |
|
52 |
### BibTex
|
53 |
```
|
|
|
3 |
---
|
4 |
# Dataset Release: Indirect Object Identification
|
5 |
`mecha_ioi` is a pair of datasets tailored for the Indirect Object Identification task, where sentences are generated from the following set of templates:
|
|
|
|
|
|
|
6 |
|
7 |
- BABA
|
8 |
```
|
|
|
45 |
"Friends {A} and {B} found a {OBJECT} at the {PLACE}. {B} gave it to {A}",
|
46 |
]
|
47 |
```
|
48 |
+
The purpose of this dataset is to facilitate interpretability research, inspired by the paper
|
49 |
+
_Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small_, from Redwood Research. We are not affiliated with Redwood Research,
|
50 |
+
and release this dataset to contribute to the collective research effort behind understanding how Transformer language models perform this task.
|
51 |
|
52 |
### BibTex
|
53 |
```
|