Update dataset card

README.md

@@ -26,7 +26,9 @@ configs:
     path: data/dev_test-*
   - split: test
     path: data/test-*
+  default: true
 dataset_info:
+  config_name: default
   features:
   - name: id
     dtype: string
@@ -40,19 +42,13 @@ dataset_info:
     sequence: string
   splits:
   - name: train
-    num_bytes: 274812860.5
     num_examples: 3500
   - name: dev
-    num_bytes: 40130069.0
     num_examples: 500
   - name: dev_test
-    num_bytes: 38926730.0
     num_examples: 500
   - name: test
-    num_bytes: 38681771.0
     num_examples: 500
-  download_size: 386301833
-  dataset_size: 392551430.5
 ---
 
 # ArGuard – Track A: Arabic Hateful Memes
@@ -86,17 +82,17 @@ Given a meme (image + Arabic text):
 | `train`      |  3,500  | full         | single-annotated bulk                   | development phase               |
 | `dev`        |    500  | full         | single-annotated bulk                   | development phase               |
 | `dev_test`   |    500  | full         | single-annotated test sample            | development phase (leaderboard); **labels released** |
-| `test`       |    500  | **dropped**  | **triple-annotated gold** (calibration) | final-evaluation phase (leaderboard) |
+| `test`       |    500  | full         | **triple-annotated gold** (calibration) | final-evaluation phase; **labels released** |
 
 - `dev_test` was the **leaderboard set** for the development phase. That
   phase is now closed and its gold labels have been **released**, so
   `dev_test` is usable as additional labelled data. It is disjoint from
   `test`.
-- `test` is the **blind test set** for the final-evaluation phase and is
-  the file you submit predictions for. It is released the same way as
-  `dev_test` — images and OCR text only, labels stripped. All 500 records
-  are triple-annotated with majority voting; the gold labels are held in
-  the CodaBench reference data and are never published here.
+- `test` was the **blind test set** for the final-evaluation phase. That
+  phase has now closed and the gold labels are **released**, so the full
+  dataset is labelled and the benchmark can be evaluated offline. All 500
+  records are triple-annotated with majority voting, which makes `test`
+  the highest-quality split and the one to report results on.
 
 ### Binary label distribution
 
@@ -105,10 +101,12 @@ Given a meme (image + Arabic text):
 | train      |  1,324  |       2,176 |     37.8% |
 | dev        |      189 |         311 |     37.8% |
 | dev_test   |      189 |         311 |     37.8% |
-| **Total (labelled)** | **1,702** | **2,798** | 37.8% |
+| test       |      148 |         352 |     29.6% |
+| **Total**  | **1,850** |   **3,150** |     37.0% |
 
-The label distribution of the blind `test` split is withheld during the
-final-evaluation phase.
+Note that `test` is deliberately **not** distribution-matched to the
+other splits: it was drawn from the triple-annotated calibration pool,
+where majority voting resolves more borderline cases to `Not Hateful`.
 
 ### Fine-grained sub-types (Subtask A2)
 
@@ -138,8 +136,8 @@ than the meme counts.
     "id": "f9a8…b1.jpg",            # str – original image filename, unique
     "image": <PIL.Image.Image>,     # embedded bytes, decoded on access
     "text": "…",                    # str – OCR-extracted Arabic meme text
-    "label": "Hateful" | "Not Hateful" | None,  # None on the blind `test` split
-    "fine_grained_label": [...],    # list[str] – empty on the blind `test` split
+    "label": "Hateful" | "Not Hateful",  # every split is now labelled
+    "fine_grained_label": [...],         # list[str], possibly empty
 }
 ```
 
@@ -155,13 +153,15 @@ train_ex = ds["train"][0]
 train_ex["image"].show()
 print(train_ex["text"], train_ex["label"], train_ex["fine_grained_label"])
 
-# `test` is the only unlabelled split — used only to produce submissions
-print(ds["test"][0]["label"])  # -> None
+test_ex = ds["test"][0]
+print(test_ex["label"], test_ex["fine_grained_label"])
 ```
 
-The blind `test` split is the submission target for the final-evaluation
-phase. Train on `train` + `dev` + `dev_test`, predict on `test`, and
-upload to CodaBench (links below).
+**Recommended benchmark protocol.** Train on `train` + `dev` + `dev_test`
+(4,500 memes, single-annotator labels) and report on `test` (500 memes,
+triple-annotated majority vote). That is the split the shared-task
+leaderboard was scored on, so results are comparable to the published
+rankings.
 
 ## Shared-task resources