Register scienceworld train_expert split in dataset card

README.md

@@ -25,6 +25,8 @@ configs:
   data_files:
   - split: train
     path: scienceworld/train.parquet
+  - split: train_expert
+    path: scienceworld/train_expert.parquet
   - split: test
     path: scienceworld/test.parquet
 ---
@@ -53,7 +55,12 @@ Every row separates the model-input field from the environment-input fields:
     `AGENT_ENV_DATA_ROOT` (default `datasets/env_assets`) and the rollout joins
     it with these relative paths; absolute paths are used as-is.
   - `expert_actions`: expert action list (non-empty for ALFWorld
-    `train_expert` / `train_hard`; used by TCOD b2f/f2b)
+    `train_expert` and ScienceWorld `train_expert`; used by TCOD b2f/f2b).
+    ALFWorld actions come from the ALFRED handcoded planner; ScienceWorld
+    actions are precomputed via the engine's built-in gold-path solver
+    (`ScienceWorldEnv.load(..., generateGoldPath=True)` +
+    `get_gold_action_sequence()`, see
+    `agent_envs/data/generate_scienceworld_expert.py`).
   - `workflow_args`: JSON string (e.g. `max_env_steps`, `mode`, `curriculum`)
   - `max_env_steps`, `mode` (`rl`/`opd`/`rl_opd`), `curriculum`
     (`none`/`b2f`/`f2b`), `split`
@@ -65,7 +72,7 @@ Load in Slime with `--input-key prompt --label-key label --metadata-key metadata
 Switch environment with the config dropdown, then pick a split:
 
 - `alfworld`: splits `train`, `train_expert`, `train_hard`, `test`, `test_unseen`
-- `scienceworld`: splits `train`, `test`
+- `scienceworld`: splits `train`, `train_expert`, `test`
 
 ## Usage (inspect a config)