Reorder dataset columns

README.md

@@ -19,6 +19,8 @@ It adds `prompt_input_token_count`, computed with `Qwen/Qwen3-4B-Base` using
 The split is repo-disjoint by `metadata.repo` and targets train/dev/test ratios
 of 0.7/0.1/0.2 while balancing the prompt token-count distribution.
 
+Column order: `id`, `prompt`, `answer`, `prompt_input_token_count`, `metadata`.
+
 | split | rows | row ratio | repos | token mean | median | p90 | p95 | max |
 |---|---:|---:|---:|---:|---:|---:|---:|---:|
 | train | 33387 | 0.7000 | 240 | 31968.8 | 15708.0 | 76027.0 | 113771.6 | 1135223 |

data/dev.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eb879bb1d09f1dac0f33372309ef459a034a155bb0f2afcda195ace8207bd4f0
-size 122858748
+oid sha256:ab1da36217b53f78d78791e1f3bc0b81d3cf2099e5811c449df551accd9231b6
+size 122826236

data/test.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:219fb73c125146f91470c558e7cb9ddece4592f95cdb91c2907954b90afdb1e8
-size 254332639
+oid sha256:62c6dc1411c4743c49cc9bafb0bdeaccec17463b1453ed4a409d22ed295526b4
+size 254235329

data/train.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d680c7e3cb7ca6aa019369d6cb7167b0907f0d918d6cc6288912abc1da814ec
-size 907523528
+oid sha256:9caa5c33ca61f025b5a79264be828bd6510101641843bebc2a31323cf3adc3ed
+size 906964596

metadata/split_summary.json

@@ -1,4 +1,11 @@
 {
+  "column_order": [
+    "id",
+    "prompt",
+    "answer",
+    "prompt_input_token_count",
+    "metadata"
+  ],
   "initial_split_objective": 3.794929162394326e-05,
   "repo_overlap_ok": true,
   "source_file": ".cache/glotcond/materialized/swe_fixer_diff_train_full.jsonl",