Simplify checkpoint variant names

README.md

@@ -25,11 +25,10 @@ included in this model repository.
 
 ```text
 checkpoints/
-  main_table/
-    <experiment>/
-      object_<object_id>/
-        handle_<part_id>/
-          policy.pth
+  <variant>/
+    object_<object_id>/
+      handle_<part_id>/
+        policy.pth
 model_manifest.jsonl
 ```
 
@@ -39,18 +38,15 @@ path, original relative path, object category, file size, and SHA-256 digest.
 
 ## Policy variants
 
-| Experiment | Variant |
+| Name | Variant |
 | --- | --- |
-| `03_state_only_ppo` | State-only PPO baseline |
-| `04_flat_history_ppo` | Flat-history PPO baseline |
-| `05_gla_no_aux` | GLA without the auxiliary objective |
-| `06_pica_no_gla_aux_v2c` | PICA without the GLA auxiliary objective (`v2c`) |
-| `07_pica_ours` | DragMesh-2 PICA policy |
-| `08_gru_ppo` | GRU PPO baseline |
-| `09_transformer_ppo` | Transformer PPO baseline |
-
-The numeric prefixes preserve the main-table experiment numbering used during
-evaluation.
+| `state` | State-only PPO baseline |
+| `history` | Flat-history PPO baseline |
+| `gla` | GLA without the auxiliary objective |
+| `pica` | PICA without the GLA auxiliary objective (`v2c`) |
+| `dragmesh2` | DragMesh-2 PICA policy |
+| `gru` | GRU PPO baseline |
+| `transformer` | Transformer PPO baseline |
 
 ## Evaluation tasks
 
@@ -86,7 +82,7 @@ import numpy as np
 import torch
 
 checkpoint_path = Path(
-    "checkpoints/main_table/07_pica_ours/"
+    "checkpoints/dragmesh2/"
     "object_45661/handle_3/policy.pth"
 )