The DeepSEA dataset is a composite benchmark extending existing public datasets
with professionally translated splits into Southeast Asian languages. While all
source code, evaluation harnesses, and scripts in the root repository are
licensed under the Apache License, Version 2.0 (../LICENSE), the dataset
splits in data/ are governed by the licenses of their upstream source
datasets:
-
IndoMMLU Subset (
indommlu.zip): Licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC-BY-NC-SA 4.0). Upstream: IndoMMLU (Koto et al., 2023). -
XQuAD Subset (
xquad.zip): Licensed under Creative Commons Attribution-ShareAlike 4.0 International (CC-BY-SA 4.0). Upstream: XQuAD (Artetxe et al., 2020). -
CrossSum Subset (
crosssum.zip): Licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC-BY-NC-SA 4.0). Upstream: CrossSum (Bhattacharjee et al., 2023). -
MGSM Subset (
mgsm.zip): Licensed under Apache License 2.0. Upstream: Multilingual Grade School Math (Shi et al., 2022). -
WMT24++ Subset (
wmt24.zip): Licensed under Creative Commons Attribution 4.0 International (CC-BY 4.0) / Apache 2.0. Upstream: WMT24 Shared Tasks.