Rebuild dataset card, provenance, composition, and Dot research identity

README.md

@@ -11,8 +11,10 @@ annotations_creators:
 - machine-generated
 source_datasets:
 - original
+thumbnail: https://huggingface.co/datasets/usedot/dot-loom-conductor-v2/resolve/main/assets/dataset-composition.png
 tags:
 - synthetic
+- structured-output
 - model-routing
 - inference-orchestration
 - multi-agent
@@ -28,15 +30,36 @@ configs:
     path: test.jsonl
 ---
 
-# Dot Loom Conductor v2
+<p align="center">
+  <img src="assets/dot-mark.png" width="104" alt="Dot">
+</p>
 
-**11,100 synthetic, empirically anchored examples for training models to emit constrained
-multi-model execution plans.** Every example describes a task, three anonymous candidate workers,
-hard call/credit/latency budgets, and the highest-utility feasible Lean, Balanced, or Strict plan.
+<h1 align="center">Dot Loom Conductor v2</h1>
 
-[View the trained adapter](https://huggingface.co/usedot/Dot-Loom-Conductor-14B) ·
-[Try the routing lab](https://huggingface.co/spaces/usedot/Dot-Loom-Conductor-Lab) ·
-[Reproduce the generator](https://github.com/usedotai/dot-loom/tree/main/research/ministral-conductor)
+<p align="center"><strong>11,100 synthetic traces for training budget-constrained multi-model routers.</strong></p>
+
+<p align="center">
+  <a href="https://huggingface.co/usedot/Dot-Loom-Conductor-14B">Trained adapter</a>
+  &nbsp;·&nbsp;
+  <a href="https://huggingface.co/spaces/usedot/Dot-Loom-Conductor-Lab">Policy explorer</a>
+  &nbsp;·&nbsp;
+  <a href="https://github.com/usedotai/dot-loom/tree/main/research/ministral-conductor">Generator and receipts</a>
+</p>
+
+Each example describes one task, three anonymous worker profiles, hard call, credit, and latency
+budgets, and the highest-utility feasible **Lean**, **Balanced**, or **Strict** execution plan.
+
+![Dataset composition](assets/dataset-composition.png)
+
+| Property | Value |
+|---|---:|
+| Examples | 11,100 |
+| Train / validation / test | 9,000 / 900 / 1,200 |
+| Policy balance | 3,700 Lean / 3,700 Balanced / 3,700 Strict |
+| Task families | 19, disjoint across splits |
+| Candidate plans evaluated per example | 15 |
+| User or production data | 0 |
+| Dataset seed | `20260716` |
 
 ## Load
 
@@ -48,22 +71,24 @@ print(dataset)
 print(dataset["train"][0]["target"])
 ```
 
-## Dataset composition
+## Split design
 
 | Split | Examples | Lean | Balanced | Strict | Task families |
 |---|---:|---:|---:|---:|---:|
 | Train | 9,000 | 3,000 | 3,000 | 3,000 | 8 |
 | Validation | 900 | 300 | 300 | 300 | 3 |
 | Test | 1,200 | 400 | 400 | 400 | 8 |
-| Total | 11,100 | 3,700 | 3,700 | 3,700 | 19 |
+| **Total** | **11,100** | **3,700** | **3,700** | **3,700** | **19** |
 
-The JSONL payload is 46,900,561 bytes. Task families are disjoint across splits. The test families
-are payment races, tenant isolation, webhook replay, OAuth integrity, SSRF egress, streaming
-settlement, health triage, and contract risk. None appears in training or validation.
+Task families are disjoint across splits. The test families are payment races, tenant isolation,
+webhook replay, OAuth integrity, SSRF egress, stream settlement, health triage, and contract risk.
+None appears in training or validation.
 
-## One row
+The JSONL payload is 46,900,561 bytes.
 
-Each row contains both the serialized completion-training pair and its inspectable source fields:
+## Schema
+
+Each row includes both a serialized completion-training pair and inspectable source fields.
 
 | Field | Type | Meaning |
 |---|---|---|
@@ -72,7 +97,7 @@ Each row contains both the serialized completion-training pair and its inspectab
 | `task` | object | Risk, complexity, consequence, ambiguity, evidence need, reversibility, and input size |
 | `workers` | list[object] | Three anonymous capability, reliability, cost, latency, provider, and strength profiles |
 | `constraints` | object | Maximum calls, credits, latency, and target minimum quality |
-| `label` | object | Lean, Balanced, or Strict plan with roles, receipts, reasons, and access graph |
+| `label` | object | Policy, roles, receipts, reasons, verification, and access graph |
 | `oracle` | object | Utility, predicted quality, pass rate, candidate count, and target attainment |
 | `family` | string | Synthetic task family used for split isolation |
 | `split` | string | Train, validation, or held-out test |
@@ -103,45 +128,48 @@ Each row contains both the serialized completion-training pair and its inspectab
 }
 ```
 
-The abbreviated example above illustrates the schema. The dataset viewer exposes every complete
-field and canonical target.
+The example is abbreviated for readability. The dataset viewer exposes every complete field and
+canonical target.
 
-## Label generation
+## Label construction
 
-For every example, the generator enumerates all 15 possible role assignments:
+For every example, the generator enumerates all 15 role assignments:
 
 - Three Lean writer-only plans
 - Six ordered writer-reviewer plans
 - Six ordered writer-reviewer-finalizer plans
 
-Plans exceeding hard call, credit, or latency limits are removed. The disclosed outcome model
-scores task-adjusted quality, pass rate, cost, latency, role fitness, independent verification,
+Plans that exceed hard call, credit, or latency budgets are rejected. The disclosed outcome model
+scores task-adjusted quality, pass rate, cost, latency, role fitness, provider independence,
 reviewer regression risk, and finalizer synthesis. The highest-utility feasible plan becomes the
-label. If any feasible plan meets the requested quality target, below-target plans are excluded.
+label. If any feasible plan meets the target quality, below-target plans are excluded.
 
 Worker profiles are sampled around three frozen empirical anchors from a six-case Dot Loom
-benchmark. Model names never appear in training prompts. Worker identifiers and provider groups
-are shuffled so the model must route from capabilities and constraints rather than brand names.
+receipt. Model names never appear in training prompts. Worker identifiers and provider groups are
+shuffled so the model must route from capability and constraint fields instead of memorizing
+brands.
 
 ## Independent label audit
 
 A blinded independent OpenAI judge audited 90 stratified train and validation labels. Test labels
 were not sent to the judge.
 
-- Agreement: 76 of 90
-- Proposed disagreements: 14
-- Feasible alternatives: 3
-- Invalid or over-budget alternatives: 11
-- Feasible alternatives improving the disclosed oracle: 0
+| Audit outcome | Count |
+|---|---:|
+| Agreement | 76 of 90 |
+| Proposed disagreements | 14 |
+| Feasible alternatives | 3 |
+| Invalid or over-budget alternatives | 11 |
+| Feasible alternatives improving the disclosed oracle | 0 |
 
-The judge is an independent critic, not ground truth. Raw audit responses and deterministic
-revalidation are available in the GitHub research package.
+The judge is an independent critic, not ground truth. Raw responses and deterministic
+revalidation are included in the public research package.
 
 ## Privacy and provenance
 
 The corpus contains no user prompts, production conversations, wallets, API keys, private code,
 or customer data. Task descriptions are synthetic. Worker identities and provider groups are
-anonymous and shuffled. Dataset seed: `20260716`.
+anonymous and shuffled.
 
 | Split | SHA-256 |
 |---|---|
@@ -152,7 +180,7 @@ anonymous and shuffled. Dataset seed: `20260716`.
 ## Intended use
 
 - Supervised fine-tuning for structured routing-plan generation
-- Research on cost, latency, verification, and model-allocation policies
+- Research on cost, latency, independence, verification, and model-allocation policies
 - Comparing learned conductors with deterministic routers
 - Testing deterministic enforcement around model-proposed plans
 
@@ -165,10 +193,7 @@ anonymous and shuffled. Dataset seed: `20260716`.
 - The balanced label distribution does not represent natural production traffic.
 - Production systems must validate learned plans with deterministic runtime enforcement.
 
-## Reproduction and citation
-
-The generator, scoring code, complete methods, raw predictions, and audit receipts are published at
-[usedotai/dot-loom](https://github.com/usedotai/dot-loom/tree/main/research/ministral-conductor).
+## Citation
 
 ```bibtex
 @dataset{dotloomconductorv2_2026,
@@ -180,3 +205,7 @@ The generator, scoring code, complete methods, raw predictions, and audit receip
   url       = {https://huggingface.co/datasets/usedot/dot-loom-conductor-v2}
 }
 ```
+
+## License
+
+Released under Apache 2.0.

SHA256SUMS

@@ -1,7 +1,9 @@
-a466c2e30628d2a15287d71cfdee69df605cff2419886f5420a8450e69465c48  ./.gitattributes
-cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  ./LICENSE
-df42c6bcc2470e26c090b0f6d890febba2782c1ac5a136edcae393ecc76e9d51  ./README.md
-53c644751b2ba45e9b03dc2b492a798dfd9c13bc056fdd52a50ba2e2f37de536  ./manifest.json
-9cd3065a36787ce18042aa8bde7e10b1ccf3acd04093058cf78c77f338c5d049  ./test.jsonl
-8e8244bb5ad7034f6d2a5949503526762895072875998662510d5a3a9b269739  ./train.jsonl
-b6579587f1123c32b58fe5e55ca9516a5e2d94c47e04a1a84405802dfa4f2795  ./validation.jsonl
+a466c2e30628d2a15287d71cfdee69df605cff2419886f5420a8450e69465c48  .gitattributes
+cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  LICENSE
+eff3ad3c85c33e46b536d82d982d1c84b7b81dfee4d2d53862f69f2742c7c0f1  README.md
+3c7fa6649e10726831731a0a06a9c017d226742666dfa9a9d4db3786a3269790  assets/dataset-composition.png
+d8b632da85953ed75b11f9e77d998ad01e89e042a14abd8f3990fc5fc2442b68  assets/dot-mark.png
+53c644751b2ba45e9b03dc2b492a798dfd9c13bc056fdd52a50ba2e2f37de536  manifest.json
+9cd3065a36787ce18042aa8bde7e10b1ccf3acd04093058cf78c77f338c5d049  test.jsonl
+8e8244bb5ad7034f6d2a5949503526762895072875998662510d5a3a9b269739  train.jsonl
+b6579587f1123c32b58fe5e55ca9516a5e2d94c47e04a1a84405802dfa4f2795  validation.jsonl