Upload 4 files

We introduce ChEBI-20-MM, a comprehensive benchmark developed from the ChEBI-20 dataset, integrating multi-modal data such as InChI, IUPAC, SELFIES, and images. This benchmark can evaluate the ability of models in molecule generation, image and IUPAC recognition, molecular captioning, and retrieval tasks.
In molecular representations, molecular text descriptions are typically scarce. MolT5's ChEBI-20 dataset in molecule generation and captioning is increased utilization. To facilitate multi-modal data evaluation, we expand our data modality variety, introducing ChEBI-20-MM.

.gitattributes

@@ -53,3 +53,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.jpg filter=lfs diff=lfs merge=lfs -text
 *.jpeg filter=lfs diff=lfs merge=lfs -text
 *.webp filter=lfs diff=lfs merge=lfs -text
+train.csv filter=lfs diff=lfs merge=lfs -text

image.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5c20d2d1721d9b9de3bbd05e8ccf641724dbad73efd7e5e851e4acf0a6c0496
+size 103056947

train.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98d4050160f325b4342d730668e3959eb7dffc3821ce1b8312b7bb9016990a00
+size 25167214