Update dataset card: add scenarios documentation

README.md

@@ -104,13 +104,23 @@ This dataset accompanies the [AutoWorldModel-Bench](https://github.com/AutoWorld
 ## Dataset Structure
 
 ```
-data/
+data/                                # Parquet training data
 └── {game}/
-    ├── train.parquet             # 10,000 episodes
-    ├── val.parquet               #  3,000 episodes
-    ├── test.parquet              #  3,000 episodes
-    ├── scenario.parquet          #  3,000 episodes
-    └── meta.json                 # max_entities, dimensions, total_frames
+    ├── train.parquet                # 10,000 episodes
+    ├── val.parquet                  #  3,000 episodes
+    ├── test.parquet                 #  3,000 episodes
+    ├── scenario.parquet             #  3,000 episodes
+    └── meta.json                    # max_entities, dimensions, total_frames
+
+scenarios/                           # Curated scenario archives (tar.gz per game)
+└── {game}.tar.gz
+    └── {game}/
+        └── {scenario_name}/         # e.g. ball_hits_paddle, ship_dies
+            └── data_ep_{id}/
+                ├── frames.jsonl.gz  # Per-frame entity states
+                ├── manifest.json    # Game schema, entity kinds, action/global fields
+                ├── meta.json        # Episode metadata, event info, rollout params
+                └── rollout.mp4      # Visual replay
 ```
 
 ## Games
@@ -141,6 +151,39 @@ v2 games (asteroids, breakout, frogger, kong, platformer) use a 3-policy mix for
 
 v1 games (pong, racer, snake) use a 50/50 heuristic + random mix.
 
+## Scenarios
+
+Hand-picked and categorized episodes that isolate specific game events — sourced from the scenario split and augmented with synthetically collected episodes. Each episode captures a short rollout around a key event (e.g., collision, scoring, death) with history context.
+
+| Game | Scenarios | Episodes | Archive Size |
+|------|---:|---:|---:|
+| asteroids | 14 | 420 | 41 MB |
+| breakout | 6 | 160 | 9 MB |
+| frogger | 5 | 180 | 27 MB |
+| kong | 5 | 180 | 7 MB |
+| platformer | 5 | 160 | 10 MB |
+| pong | 15 | 460 | 14 MB |
+| racer | 5 | 140 | 6 MB |
+| snake | 5 | 160 | 11 MB |
+
+Each game includes a `same_state_different_actions` scenario that tests action-conditioning by replaying the same initial state with varied actions.
+
+### Downloading scenarios
+
+```python
+from huggingface_hub import hf_hub_download
+import tarfile
+
+path = hf_hub_download(
+    "AutoWorldModel/AutoWorldModelBench",
+    "scenarios/pong.tar.gz",
+    repo_type="dataset",
+)
+with tarfile.open(path) as tar:
+    tar.extractall("./scenarios")
+# ./scenarios/pong/ball_hits_left_paddle_moving/data_ep_.../frames.jsonl.gz
+```
+
 ## Tensor Schema
 
 Each Parquet row stores one episode as serialized numpy arrays: