Use per-task train/test configs (data untouched)

README.md

@@ -7,167 +7,205 @@ tags:
 - role-play
 - social-intelligence
 configs:
-  - config_name: alignx_arbitrary
+  - config_name: alignx_train
     default: true
     data_files:
       - split: train
         path: train/alignx_rl_8k.parquet
+  - config_name: behaviorchain_train
+    data_files:
+      - split: train
+        path: train/behaviorchain_rl_train.parquet
+  - config_name: coser_train
+    data_files:
+      - split: train
+        path: train/coser_rl_train.parquet
+  - config_name: fantom_train
+    data_files:
+      - split: train
+        path: train/fantom_rl_train.parquet
+  - config_name: hitom_train
+    data_files:
+      - split: train
+        path: train/hitom_rl_train.parquet
+  - config_name: humanllm_train
+    data_files:
+      - split: train
+        path: train/humanllm_rl_train.parquet
+  - config_name: humanual_book_train
+    data_files:
+      - split: train
+        path: train/humanual_rl_book.parquet
+  - config_name: humanual_chat_train
+    data_files:
+      - split: train
+        path: train/humanual_rl_chat.parquet
+  - config_name: humanual_email_train
+    data_files:
+      - split: train
+        path: train/humanual_rl_email.parquet
+  - config_name: humanual_news_train
+    data_files:
+      - split: train
+        path: train/humanual_rl_news.parquet
+  - config_name: humanual_opinion_train
+    data_files:
+      - split: train
+        path: train/humanual_rl_opinion.parquet
+  - config_name: humanual_politics_train
+    data_files:
+      - split: train
+        path: train/humanual_rl_politics.parquet
+  - config_name: lifechoices_train
+    data_files:
+      - split: train
+        path: train/lifechoices_hard_rl.parquet
+  - config_name: mirrorbench_train
+    data_files:
+      - split: train
+        path: train/mirrorbench_rl_train.parquet
+  - config_name: mistakes_train
+    data_files:
+      - split: train
+        path: train/mistakes_rl_train.parquet
+  - config_name: paratomi_train
+    data_files:
+      - split: train
+        path: train/paratomi_rl_train.parquet
+  - config_name: sim_doc_train
+    data_files:
+      - split: train
+        path: train/sim_doc_rl.parquet
+  - config_name: sim_math_train
+    data_files:
+      - split: train
+        path: train/sim_math_rl.parquet
+  - config_name: social_r1_train
+    data_files:
+      - split: train
+        path: train/social_r1_rl.parquet
+  - config_name: socsci210_train
+    data_files:
+      - split: train
+        path: train/socsci210_rl_2k.parquet
+  - config_name: sotopia_train
+    data_files:
+      - split: train
+        path: train/sotopia_clean_rl.parquet
+  - config_name: twinvoice_train
+    data_files:
+      - split: train
+        path: train/twinvoice_rl_train.parquet
+  - config_name: userllm_train
+    data_files:
+      - split: train
+        path: train/userllm_rl_train.parquet
+  - config_name: alignx_arbitrary_test
+    data_files:
       - split: test
         path: test/alignx_arbitrary_val.parquet
-  - config_name: alignx_demo
+  - config_name: alignx_demo_test
     data_files:
-      - split: train
-        path: train/alignx_rl_8k.parquet
       - split: test
         path: test/alignx_demo_val.parquet
-  - config_name: alignx_history16
+  - config_name: alignx_history16_test
     data_files:
-      - split: train
-        path: train/alignx_rl_8k.parquet
       - split: test
         path: test/alignx_history16_val.parquet
-  - config_name: alignx_pair
+  - config_name: alignx_pair_test
     data_files:
-      - split: train
-        path: train/alignx_rl_8k.parquet
       - split: test
         path: test/alignx_pair_val.parquet
-  - config_name: alignx_ugc
+  - config_name: alignx_ugc_test
     data_files:
-      - split: train
-        path: train/alignx_rl_8k.parquet
       - split: test
         path: test/alignx_ugc_val.parquet
-  - config_name: behaviorchain
+  - config_name: behaviorchain_test
     data_files:
-      - split: train
-        path: train/behaviorchain_rl_train.parquet
       - split: test
         path: test/behaviorchain_val.parquet
-  - config_name: coser
+  - config_name: coser_test
     data_files:
-      - split: train
-        path: train/coser_rl_train.parquet
       - split: test
         path: test/coser_val.parquet
-  - config_name: fantom
+  - config_name: fantom_test
     data_files:
-      - split: train
-        path: train/fantom_rl_train.parquet
       - split: test
         path: test/fantom_val.parquet
-  - config_name: hitom
+  - config_name: hitom_test
     data_files:
-      - split: train
-        path: train/hitom_rl_train.parquet
       - split: test
         path: test/hitom_val.parquet
-  - config_name: humanllm
+  - config_name: humanllm_test
     data_files:
-      - split: train
-        path: train/humanllm_rl_train.parquet
       - split: test
         path: test/humanllm_val.parquet
-  - config_name: humanual_book
+  - config_name: humanual_book_test
     data_files:
-      - split: train
-        path: train/humanual_rl_book.parquet
       - split: test
         path: test/humanual_book_val.parquet
-  - config_name: humanual_chat
+  - config_name: humanual_chat_test
     data_files:
-      - split: train
-        path: train/humanual_rl_chat.parquet
       - split: test
         path: test/humanual_chat_val.parquet
-  - config_name: humanual_email
+  - config_name: humanual_email_test
     data_files:
-      - split: train
-        path: train/humanual_rl_email.parquet
       - split: test
         path: test/humanual_email_val.parquet
-  - config_name: humanual_news
+  - config_name: humanual_news_test
     data_files:
-      - split: train
-        path: train/humanual_rl_news.parquet
       - split: test
         path: test/humanual_news_val.parquet
-  - config_name: humanual_opinion
+  - config_name: humanual_opinion_test
     data_files:
-      - split: train
-        path: train/humanual_rl_opinion.parquet
       - split: test
         path: test/humanual_opinion_val.parquet
-  - config_name: humanual_politics
+  - config_name: humanual_politics_test
     data_files:
-      - split: train
-        path: train/humanual_rl_politics.parquet
       - split: test
         path: test/humanual_politics_val.parquet
-  - config_name: lifechoices
+  - config_name: lifechoices_test
     data_files:
-      - split: train
-        path: train/lifechoices_hard_rl.parquet
       - split: test
         path: test/lifechoices_val.parquet
-  - config_name: mirrorbench
+  - config_name: mirrorbench_test
     data_files:
-      - split: train
-        path: train/mirrorbench_rl_train.parquet
       - split: test
         path: test/mirrorbench_val.parquet
-  - config_name: mistakes
+  - config_name: mistakes_test
     data_files:
-      - split: train
-        path: train/mistakes_rl_train.parquet
       - split: test
         path: test/mistakes_val.parquet
-  - config_name: paratomi
+  - config_name: paratomi_test
     data_files:
-      - split: train
-        path: train/paratomi_rl_train.parquet
       - split: test
         path: test/paratomi_val.parquet
-  - config_name: sim_doc
+  - config_name: sim_doc_test
     data_files:
-      - split: train
-        path: train/sim_doc_rl.parquet
       - split: test
         path: test/sim_doc_val.parquet
-  - config_name: sim_math
+  - config_name: sim_math_test
     data_files:
-      - split: train
-        path: train/sim_math_rl.parquet
       - split: test
         path: test/sim_math_val.parquet
-  - config_name: social_r1
+  - config_name: social_r1_test
     data_files:
-      - split: train
-        path: train/social_r1_rl.parquet
       - split: test
         path: test/social_r1_val.parquet
-  - config_name: socsci210
+  - config_name: socsci210_test
     data_files:
-      - split: train
-        path: train/socsci210_rl_2k.parquet
       - split: test
         path: test/socsci210_val.parquet
-  - config_name: sotopia
+  - config_name: sotopia_test
     data_files:
-      - split: train
-        path: train/sotopia_clean_rl.parquet
       - split: test
         path: test/sotopia_hard_val.parquet
-  - config_name: twinvoice
+  - config_name: twinvoice_test
     data_files:
-      - split: train
-        path: train/twinvoice_rl_train.parquet
       - split: test
         path: test/twinvoice_val.parquet
-  - config_name: userllm
+  - config_name: userllm_test
     data_files:
-      - split: train
-        path: train/userllm_rl_train.parquet
       - split: test
         path: test/userllm_val.parquet
 ---
@@ -182,16 +220,25 @@ spanning conversation, social simulation, social cognition, role-play, and human
 - 💻 Code: https://github.com/sunnweiwei/OdysSim
 - 🤖 Model: https://huggingface.co/sunweiwei/Ditto-8B
 
-Each task is exposed as a separate config with a `train` (RL) and `test` (evaluation) split:
+## Structure
+
+Files are kept in their original format, bound to our training/eval framework, and are organized
+into two folders:
+
+- `train/` — RL training data, one parquet per task.
+- `test/` — evaluation data, one parquet per task.
+
+Because each task's `extra_info` schema differs across tasks and between train/eval, each
+task+split is exposed as its own config (`<task>_train`, `<task>_test`) so it loads cleanly:
 
 ```python
 from datasets import load_dataset
-ds = load_dataset("sunweiwei/Soul", "userllm")
-ds["train"], ds["test"]
+train = load_dataset("sunweiwei/Soul", "userllm_train", split="train")
+test  = load_dataset("sunweiwei/Soul", "userllm_test",  split="test")
 ```
 
-Files are kept in their original format (bound to our training/eval framework); the configs
-above only map each task's RL/eval parquet to a `train`/`test` split.
+You can also read any parquet directly, e.g. `train/userllm_rl_train.parquet` /
+`test/userllm_val.parquet`.
 
 ## Citation