Update README.md

README.md

@@ -68,17 +68,17 @@ Rather than training one model, we trained multiple models, each on a different
 Accuracy is normalized by length and shown as a percentage.
 | Dataset                                                 | ARC-Easy ↑ | ARC-Challenge ↑ | HellaSwag ↑ | PIQA ↑ | SWAG ↑ | BLiMP ↑ |  Avg ↑ |   **II ↑** |
 | :------------------------------------------------------ | ---------: | --------------: | ----------: | -----: | -----: | ------: | -----: | ---------: |
-| [FineWeb](https://huggingface.co/HuggingFaceFW/fineweb) |     27.82% |          22.01% |      26.89% | 52.83% | 30.34% |  65.21% | 37.52% | **39.33%** |
-| DCLM-1.0-Baseline                                       |     29.08% |          22.41% |      27.06% | 52.23% | 31.00% |  66.16% | 37.99% | **39.82%** |
-| DOAB                                                    |     31.14% |          20.99% |      28.00% | 52.18% | 29.43% |  62.15% | 37.32% | **39.10%** |
+| [FineWeb](https://huggingface.co/HuggingFaceFW/fineweb) |     27.82% |          22.01% |      26.89% | 52.83% | 30.34% |  65.21% | 37.52% | **37.49%** |
+| DCLM-1.0-Baseline                                       |     29.08% |          22.41% |      27.06% | 52.23% | 31.00% |  66.16% | 37.99% | **37.92%** |
+| DOAB                                                    |     31.14% |          20.99% |      28.00% | 52.18% | 29.43% |  62.15% | 37.32% | **37.61%%** |
 
 The II is calculated based on these weights:
 
-| Benchmark     | Weight | Note (Why?)                                                                                      |
+| Benchmark     | Weight | Note (Why?)                                                                                 |
 | :------------ | -----: | :------------------------------------------------------------------------------------------ |
 | ARC-Easy      |   1.00 | —                                                                                           |
-| ARC-Challenge |   0.65 | ARC-Challenge is consistently below chance at small scales, so we assign it a lower weight. |
+| ARC-Challenge |   0.5  | ARC-Challenge is consistently below chance at small scales, so we assign it a much lower weight. |
 | HellaSwag     |   1.00 | —                                                                                           |
-| SWAG          |   0.95 | SWAG is an easier version of HellaSwag, so we assign it a slightly lower weight.            |
+| SWAG          |   0.75 | SWAG is an easier version of HellaSwag, so we assign it a lower weight.            |
 | PIQA          |   1.00 | —                                                                                           |
 | BLiMP         |   0.65 | BLiMP primarily measures grammatical correctness rather than broader reasoning ability.     |