Update dataset usage wording
Browse files
README.md
CHANGED
|
@@ -19,8 +19,8 @@ theorem-proving tasks with bundled oracle solutions. The executable benchmark is
|
|
| 19 |
distributed as a Harbor dataset; this Hugging Face dataset is the metadata and
|
| 20 |
discovery mirror.
|
| 21 |
|
| 22 |
-
The benchmark is intended for evaluation
|
| 23 |
-
|
| 24 |
|
| 25 |
## Running
|
| 26 |
|
|
|
|
| 19 |
distributed as a Harbor dataset; this Hugging Face dataset is the metadata and
|
| 20 |
discovery mirror.
|
| 21 |
|
| 22 |
+
The benchmark is intended for evaluation or RL. Benchmark target files and
|
| 23 |
+
oracle solutions should not be included in training corpora.
|
| 24 |
|
| 25 |
## Running
|
| 26 |
|