Update dataset

.gitattributes

@@ -57,3 +57,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+train.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,12 +1,13 @@
-# Rush Hour 4×4, 5×5 & 6×6 — Balanced (solvable-only, A present)
-Enumerates boards (4×4/5×5), samples for selected sizes (e.g., 6×6), BFS-solves to optimal sequences, and stratifies by difficulty.
+# Rush Hour 4×4, 5×5 & 6×6 — Balanced (solvable-only, fixed thresholds)
+Enumerates boards (4×4/5×5) and/or samples for selected sizes (e.g., 6×6), BFS-solves to optimal sequences, and stratifies by difficulty.
 
 ## Policies
-- sizes: 5, 6
+- sizes: 4, 5, 6
 - solvable_only: True
 - exclude_already_solved: True
-- min_moves: 2
-- difficulty_quantiles: 0.33,0.66 (per size)
+- difficulty_scheme: fixed
+- fixed_thresholds: 3,5  # easy < t1, medium < t2, hard ≥ t2
+- balance_per_size: True
 
 ## Schema
 - board (str) | size (int) | moves (int) | solution (list[str]) | difficulty (easy/medium/hard) | messages (chat)

train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08892f4a342379ef40aa95aefd693793e48cc26fa04e642eafb12d7f82e6181a
+size 43767154