benchmark-only: single test split of 400; add measured score reference points

README.md

@@ -17,42 +17,22 @@ tags:
 configs:
 - config_name: customer_service
   data_files:
-  - split: train
-    path: customer_service/train-*.parquet
-  - split: validation
-    path: customer_service/validation-*.parquet
   - split: test
     path: customer_service/test-*.parquet
 - config_name: security_incidents
   data_files:
-  - split: train
-    path: security_incidents/train-*.parquet
-  - split: validation
-    path: security_incidents/validation-*.parquet
   - split: test
     path: security_incidents/test-*.parquet
 - config_name: invoice_processing
   data_files:
-  - split: train
-    path: invoice_processing/train-*.parquet
-  - split: validation
-    path: invoice_processing/validation-*.parquet
   - split: test
     path: invoice_processing/test-*.parquet
 - config_name: agent_trace_observability
   data_files:
-  - split: train
-    path: agent_trace_observability/train-*.parquet
-  - split: validation
-    path: agent_trace_observability/validation-*.parquet
   - split: test
     path: agent_trace_observability/test-*.parquet
 - config_name: all
   data_files:
-  - split: train
-    path: all/train-*.parquet
-  - split: validation
-    path: all/validation-*.parquet
   - split: test
     path: all/test-*.parquet
 ---
@@ -158,6 +138,31 @@ accuracy.
    what makes the gold soft where the decision is genuinely ambiguous.
 4. **Audited before release** — see `AUDIT.md`.
 
+## What a score here means
+
+Gold is the mean of three samples from a ~4B-class teacher endpoint, so a score
+measures **agreement with that teacher**, not correctness. Three reference points,
+all measured on these 400 cases:
+
+| Reference | Accuracy | What it is |
+|---|---|---|
+| Majority baseline | 0.520 | ignore the input, always guess the commonest label |
+| Perfect scenario understanding | 0.704 | a model fitted to the latent factors that generated each case, cross-validated |
+| Teacher self-agreement | 0.735 | a fresh teacher sample scored against gold built from the others |
+
+**Read 0.52 as the floor, ~0.70 as strong, ~0.75 as saturation.** The factor
+ceiling sits *below* teacher self-agreement because the teacher shares its own
+idiosyncrasies with the gold and an outside model does not. A score much above
+0.75 means a model has learned the teacher's quirks rather than the task.
+
+Per-question ceilings vary widely -- from 0.560 on `agent_trace/urgency` to
+0.937 on `customer_service/category` -- so every score should be read against
+its own question, not against the mean.
+
+A better model can score *worse*: anything right where the teacher is wrong is
+penalised. The teacher missed a duplicate invoice whose ID literally matched a
+prior one.
+
 ## Honest limitations
 
 **The gold is a teacher signal, not ground truth.** It comes from a model of
@@ -173,13 +178,14 @@ text, not drawn from real support queues, SIEMs or accounts-payable systems.
 Results here describe architecture and systems behaviour, not real-world
 accuracy.
 
-**It is small.** 400 cases, 2000 typed decisions. Enough for
-decision-level metrics; thin for strong per-workflow claims. The generator is
-seeded and extensible, so the dataset can grow without invalidating these cases.
+**It is small.** 400 cases, 2000 typed decisions -- all of them
+held out. Enough for decision-level metrics; thin for strong per-workflow claims.
+The generator is seeded and extensible, so the benchmark can grow without
+invalidating these cases.
 
-**Splits are dealt by case index, not stratified by label.** Assignment is
-exactly balanced in size and fully reproducible, but per-split *label* balance is
-whatever chance gives.
+**It is a benchmark, not a training corpus.** Every case is a held-out test
+case. Generate training or calibration data separately, with a different seed;
+do not carve a train split out of this one.
 
 ## What the audit checks
 

agent_trace_observability/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8657f0565dd8b7df91c300c2423e83c358d6eef85a3cfde0d06cbd55ccca6a1d
-size 33838
+oid sha256:833ad1364237b9a3fe05be2c62e632f0cb34ba372284b975140d362ab246f70d
+size 71362

agent_trace_observability/train-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:17aec045c98eda1cc48dd76300c88ba68772db86bb7efdc14528480b9cbaacec
-size 58131

agent_trace_observability/validation-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:029693f27f46e1afd0feef8aedae96095fc6a37c2d9e49c2a14e5ca24db6c1e7
-size 33454

all/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9fd727b84084b72337fead6179679b10c878fde87bd35148b3bf7049a1377264
-size 57848
+oid sha256:6c777b77cb1b45fd85ed83d1bdfdb328bb728b2bb7d05170a34a42b5f4228565
+size 222129

all/train-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d713af9dc7e24c393ce17d523724c11d114ba8bacc9a3df4fa05ca8b97b36a5c
-size 164208

all/validation-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4082612f494392a0e1aa10066d226f4b64d4307da6519f8bf206014393068e70
-size 57493

customer_service/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf63e27320f6183869ba8b61b1bb37837ae5a0aaad28e4d16391dad8d27d95e6
-size 45545
+oid sha256:b4e6e23b3c2add197dda74008f2ce42ff996511ff1bf9c93e8b0e6d84ac9c275
+size 115037

customer_service/train-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:fa2f6eba9ee6c121a48e5a662f1e1a82c5aa9fe8cfae1d8c16a0c8cd3b35c083
-size 89774

customer_service/validation-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b1231df615d84efd588a15310f30a7f3c5143ae517fe750c661649d36307201d
-size 44516

invoice_processing/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06ebe11dd258de57198a3ebd234b274a8acbaeff9b8868b8a1dcf5987d53568c
-size 35725
+oid sha256:ce2adfb5201dba910b23bda6ab9678c385b580edaabbf8348b4e024c120d4b75
+size 79603

invoice_processing/train-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:dd9a62ae2f5622ee707f4a9c0e526a518cdc508ebd2867d40db68a0f737abedf
-size 64073

invoice_processing/validation-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:992fc8e6244c3b867b68410a629e80b464d2696e65c25f625645c5af9dd9caf6
-size 35538

security_incidents/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b23dfd868ca75214f24bf6a16a0022e10f7214dd7fb882049af4e6cd6dd4c0bf
-size 37141
+oid sha256:d9dcda135003e6812bec87461fcf53591e8b5426b077836616cde871d26f8dda
+size 87540

security_incidents/train-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ffaeeafc23053ae896ed575b04d570751305938c38b101a44bc557b8ac6e6a89
-size 69884

security_incidents/validation-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f45915cd05e65c5b63467d034fdc72b692faf4df8de971b4cb6a75de0fa344fb
-size 37736