Ashish Rai
commited on
Commit
·
a2813c9
1
Parent(s):
bb52b99
Add results
Browse files
README.md
CHANGED
|
@@ -47,6 +47,12 @@ The dataset has been used to evaluate several state-of-the-art models, including
|
|
| 47 |
- Accuracy declines sharply for operations involving larger numbers or higher unique digit counts.
|
| 48 |
- Models exhibit higher accuracy for the first and last digits of predictions compared to middle digits.
|
| 49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
## Usage
|
| 51 |
|
| 52 |
### Loading the Dataset
|
|
|
|
| 47 |
- Accuracy declines sharply for operations involving larger numbers or higher unique digit counts.
|
| 48 |
- Models exhibit higher accuracy for the first and last digits of predictions compared to middle digits.
|
| 49 |
|
| 50 |
+

|
| 51 |
+

|
| 52 |
+

|
| 53 |
+

|
| 54 |
+

|
| 55 |
+
|
| 56 |
## Usage
|
| 57 |
|
| 58 |
### Loading the Dataset
|
assets/add-tuned-lens-llama-3.1-8b-2.png
ADDED
|
Git LFS Details
|
assets/add-tuned-lens-llama-3.1-8b.png
ADDED
|
Git LFS Details
|
assets/length-generalization-accuracy.png
ADDED
|
Git LFS Details
|
assets/mul-digit-accuracy-gpt-4o-mini.png
ADDED
|
Git LFS Details
|
assets/mul-digit-accuracy-llama-3.1-70b.png
ADDED
|
Git LFS Details
|