Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -211,7 +211,7 @@ Analysis revealed that FineWeb's C4 filter removes pages containing '{' characte
|
|
| 211 |
The expanded corpus underwent a more fine-grained quality evaluation:
|
| 212 |
|
| 213 |
Once again, we used LLama-3.1-70B-Instruct to score a sample of newly extracted pages on a 5-point scale (full prompt available in [here](prompt.txt)):
|
| 214 |
-
We finetuned a new classifier
|
| 215 |
After leaving only pages with a score of 3 or higher, and deduplicating the samples using simple single-band MinHash-LSH, we obtained FineMath-3+ with 34B tokens.
|
| 216 |
|
| 217 |
The same classifier was applied to InfiMM-WebMath's text content, focusing more on reasoning rather than advanced mathematics.
|
|
|
|
| 211 |
The expanded corpus underwent a more fine-grained quality evaluation:
|
| 212 |
|
| 213 |
Once again, we used LLama-3.1-70B-Instruct to score a sample of newly extracted pages on a 5-point scale (full prompt available in [here](prompt.txt)):
|
| 214 |
+
We finetuned a new [classifier](https://huggingface.co/HuggingFaceTB/finemath-classifier) on these annotations and scored the entire corpus.
|
| 215 |
After leaving only pages with a score of 3 or higher, and deduplicating the samples using simple single-band MinHash-LSH, we obtained FineMath-3+ with 34B tokens.
|
| 216 |
|
| 217 |
The same classifier was applied to InfiMM-WebMath's text content, focusing more on reasoning rather than advanced mathematics.
|