Document grounded-max-100m release

README.md

@@ -32,13 +32,19 @@ configs:
     path: "data/dense-max-100m/train/*"
   - split: validation
     path: "data/dense-max-100m/validation/*"
+- config_name: grounded-max-100m
+  data_files:
+  - split: train
+    path: "data/grounded-max-100m/train/*"
+  - split: validation
+    path: "data/grounded-max-100m/validation/*"
 ---
 
 <div align="center">
 
 # StateBench v1
 
-### 210 million verified state episodes for pretraining, SFT, and on-policy SDFT
+### 310 million verified state episodes for pretraining, SFT, and on-policy SDFT
 
 Train models to **retain**, **edit**, **address**, and **transform** state across long execution traces.
 
@@ -51,8 +57,9 @@ Train models to **retain**, **edit**, **address**, and **transform** state acros
 | **default** | 10,000,000 programs · 7.80 GiB | You need 36 explicit families and varied surfaces |
 | **dense-100m** | 100M episodes · 94.78 GiB | You need high-throughput training on a compact shared VM |
 | **dense-max-100m** | 100M episodes · 93.45 GiB | You need maximum-density stress training on the compact register VM |
+| **grounded-max-100m** | 100M episodes · 96.57 GiB | You want the dense-max oracles in domain-grounded, structurally varied prompts |
 
-Both configurations contain structured prompt, completion, and teacher-only privileged context. Every target is computed by an executable reference machine, never authored by a teacher model.
+All configurations contain structured prompt, completion, and teacher-only privileged context. Every target is computed by an executable reference machine, never authored by a teacher model.
 
 ## Load
 
@@ -67,12 +74,16 @@ Both configurations contain structured prompt, completion, and teacher-only priv
     dense_max = load_dataset(
         "aabbdev/StateBench-v1", "dense-max-100m", split="train", streaming=True
     )
+    grounded = load_dataset(
+        "aabbdev/StateBench-v1", "grounded-max-100m", split="train", streaming=True
+    )
 
 | Configuration | Train | Validation |
 |:--|--:|--:|
 | default | 9,949,600 rows | 50,400 rows |
 | dense-100m | 24,875,000 rows | 125,000 rows |
 | dense-max-100m | 24,875,000 rows | 125,000 rows |
+| grounded-max-100m | 24,875,000 rows | 125,000 rows |
 
 ## Train three ways
 
@@ -88,7 +99,7 @@ privileged_context is visible only to the teacher. Never concatenate it into the
 
 ### Causal pretraining
 
-The default configuration includes a ready-to-use text column. dense-100m avoids storing a duplicate ~95 GiB text view; derive it while streaming:
+The default configuration includes a ready-to-use text column. Dense configurations avoid storing a duplicate text view; derive it while streaming:
 
     def to_text(row):
         prompt = row["prompt"][0]["content"]
@@ -115,7 +126,9 @@ The default configuration includes a ready-to-use text column. dense-100m avoids
 | State systems | Transactional + temporal · Adaptive + control · Concurrent + distributed · Graph + structures |
 | Stress systems | Robust streaming · Multi-agent state · Algebraic automata · Memory pressure |
 
-default implements 36 explicit behavioral families. dense-100m and dense-max-100m are compact register-VM curricula. dense-max-100m balances 384 curriculum labels and a pairwise schedule of 32 renderer IDs and 64 transformation IDs; those IDs describe coverage scheduling, while its published prompt surface remains intentionally VM-centric. The separately qualified 40-opcode/384-mechanism compiler is the foundation for the forthcoming grounded configuration.
+default implements 36 explicit behavioral families. dense-100m and dense-max-100m are compact register-VM curricula. dense-max-100m balances 384 curriculum labels and a pairwise schedule of 32 renderer IDs and 64 transformation IDs; those IDs describe coverage scheduling, while its published prompt surface remains intentionally VM-centric.
+
+grounded-max-100m rematerializes the same verified dense-max trajectories through 24 domain dialects: the 16 primitive domains and 8 compositions above. Its 32 document structures and 64 metamorphic surface passes are physically present in the prompts, including data formats, code-like records, logs, traces, conversations, and nested documents. This configuration targets surface transfer while preserving the exact audited answers; it remains synthetic state training, not a factual natural-language corpus.
 
 ## Expected model capabilities
 
@@ -150,15 +163,16 @@ StateBench trains state handling, not factual knowledge. It does not by itself e
 |:--|--:|--:|
 | dense-100m | 162.01 units / 1k chars | 107–120B tokens |
 | dense-max-100m | 162.88 units / 1k chars | ~107–120B tokens |
+| grounded-max-100m | 160.61 units / 1k chars | ~107–120B tokens |
 
 The token range was measured on 1,000 real packed rows with GPT-2 and Qwen3 tokenizers. Actual training cost depends on chat templates, truncation, packing, and the SDFT generation policy.
 
 ## Integrity
 
-- 210M semantic identities are unique across their respective releases.
+- 310M semantic identities are unique within their respective configurations.
 - Required fields, roles, answer JSON, and SDFT isolation passed full-corpus audits.
 - No benchmark name, family label, or privileged answer appears in model-visible prompts.
 - SHA-256 checksums cover every Parquet shard.
 - The public training corpus is separate from private StateBench evaluation material.
 
-Release manifests: dataset_manifest.json, dense-100m-manifest.json, and dense-max-100m-manifest.json.
+Release manifests: dataset_manifest.json, dense-100m-manifest.json, dense-max-100m-manifest.json, and grounded-max-100m-manifest.json.