Dataset Viewer
Auto-converted to Parquet Duplicate
chapter_id
stringlengths
16
16
family
stringclasses
6 values
template_families
listlengths
1
4
target_topic_id
int32
1
199
ablation
stringclasses
1 value
template_ids
listlengths
1
5
style_topic_ids
listlengths
3
3
model
stringclasses
2 values
prompt_kind
stringclasses
2 values
temperature
float32
0.4
0.4
seed
int32
20.3M
20.3M
prompt_chars
int32
5.28k
9.79k
response_chars
int32
0
18.9k
response_text
stringlengths
0
18.9k
response_reasoning
stringlengths
144
33.3k
hx_exchange_id
stringlengths
36
36
latency_ms
int32
6.14k
52.3k
prompt_tokens
int32
1.29k
2.46k
completion_tokens
int32
1.52k
8.19k
reasoning_tokens
int32
438
8.19k
cost_usd
float32
0.01
0.03
created_at
timestamp[us, tz=UTC]date
2026-06-07 22:59:38
2026-06-08 21:13:02
audit_run_id
stringclasses
1 value
8fc927a681ee1c82
03_directive_governance
[ "03_directive_governance", "07_long_tail" ]
71
full
[ "iso27001_annexa_subclass", "attestation_with_supporting_evidence", "iso27001_clause_subclass", "verification_produces_evidence", "requirement_equiv_specifies_target" ]
[ 71, 146, 2 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,608
8,201
14,862
# Governance Frameworks: ISO/IEC 27001 Compliance and Evidence Verification ## Scope and Preamble This chapter delineates the formal ontological structures underpinning information security management systems (ISMS), specifically focusing on the integration of ISO/IEC 27001 controls, clause requirements, and the evid...
1. **Analyze the Request:** * **Goal:** Write one chapter of a technical textbook. * **Topic:** A small number of related concepts based on provided OWL ontology axioms. * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance frameworks). Dense, evidence-anc...
435ed0cf-606e-4d53-bd92-7c3be6aece71
11,590
2,146
6,844
3,786
0.02365
2026-06-08T02:20:20.583000
232ea5460ce6e0bf
b2ba7aa53fe4d237
07_long_tail
[ "03_directive_governance", "07_long_tail" ]
199
full
[ "schemaorg_creativework_date", "requirement_equiv_specifies_target", "schemaorg_event_starttime", "schemaorg_event_endtime" ]
[ 199, 155, 7 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,609
7,540
11,835
# Chapter 4: Governance of Artifacts and Directives in Compliance Frameworks ## 4.1 Scope and Preamble This chapter delineates the structural relationships governing the creation, publication, and temporal execution of compliance artifacts within a regulated environment. It establishes the formal ontology required to...
1. **Analyze the Request:** * **Goal:** Write a single chapter of a technical textbook. * **Topic:** Teach a small number of related concepts precisely (based on provided OWL ontology axioms). * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance framework...
6b2c13f2-8993-4d33-a1d8-83e7f8917f30
10,014
2,010
6,228
3,563
0.02165
2026-06-08T02:20:45.282000
232ea5460ce6e0bf
c8c05acfc5668f7d
03_directive_governance
[ "03_directive_governance" ]
75
full
[ "policy_equiv_jurisdiction_and_scope", "attestation_signed_by", "policy_min_one_enforcer", "requirement_equiv_specifies_target", "attestation_min_one_signer" ]
[ 75, 64, 128 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,610
7,768
11,135
# Chapter 4: Governance Structures in Sustainable Infrastructure Compliance This chapter delineates the formal governance structures necessary for the enforcement of sustainable infrastructure standards, specifically within the context of municipal and regulatory frameworks. It provides a rigorous examination of direc...
1. **Analyze the Request:** * **Goal:** Write a technical textbook chapter. * **Topic:** A small number of related concepts based on OWL ontology axioms provided. * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance frameworks). Dense, evidence-anchored, ...
07217d94-2370-443f-b638-4d6c57f497d1
7,572
2,073
4,625
2,162
0.017383
2026-06-08T02:21:07.352000
232ea5460ce6e0bf
22e4ad24193ca971
07_long_tail
[ "03_directive_governance", "07_long_tail" ]
13
full
[ "control_only_addresses_threat_class", "hipaa_safeguard_physical", "hipaa_safeguard_admin", "hipaa_safeguard_technical" ]
[ 13, 176, 47 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,611
7,456
11,378
# Chapter 7: Governance Frameworks for Risk Addressal and Safeguard Implementation ## 7.1 Scope and Applicability This chapter establishes the mandatory directives for the classification, implementation, and auditing of risk control measures within the enterprise information security environment. The content herein d...
1. **Analyze the Request:** * **Goal:** Write one chapter of a technical textbook. * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance frameworks). Dense, evidence-anchored prose. Match the register of the provided style references (SFPE dues notice, RPU Boar...
ba15f353-65eb-4cb4-9fb1-7916ed729e16
10,623
2,024
6,549
3,943
0.022276
2026-06-08T02:21:32.550000
232ea5460ce6e0bf
6b9de69f513dc49b
04_ebpf_kernel
[ "04_ebpf_kernel" ]
64
full
[ "ebpfprogram_verified_safe" ]
[ 64, 76, 46 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,612
5,449
11,539
# Chapter 7: Governance of Kernel Space Artifacts and Verification Protocols ## 7.1 Scope and Preamble This chapter delineates the formal governance framework governing the lifecycle management of extended Berkeley Packet Filter (eBPF) programs within controlled computing environments. It establishes the ontological ...
1. **Analyze the Request:** * **Goal:** Write one chapter of a technical textbook (1500-3000 words). * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance frameworks, regulatory guides). Dense, evidence-anchored prose. High-quality PDF style. * **Style Ref...
ce150cdb-fb2f-4020-b97c-e42eda50703c
10,645
1,401
5,200
2,846
0.017164
2026-06-08T02:21:57.628000
232ea5460ce6e0bf
ca37329051fef908
04_ebpf_kernel
[ "04_ebpf_kernel" ]
46
full
[ "syscall_audited_by" ]
[ 46, 188, 54 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,613
5,411
9,820
# Chapter 4: Ontological Mapping of eBPF Audit Provenance ## 4.1 Scope and Preamble This chapter delineates the formal ontological structures governing kernel-level syscall auditing within extended Berkeley Packet Filter (eBPF) environments. It establishes the requisite definitions for designative information content...
1. **Analyze the Request:** * **Goal:** Write a single chapter of a technical textbook. * **Topic:** A small number of related concepts based on provided OWL ontology axioms. * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance frameworks, regulatory guid...
311f9c21-1d19-4410-8c7a-55daf7783d8c
9,733
1,399
5,264
3,109
0.017336
2026-06-08T02:22:21.928000
232ea5460ce6e0bf
851314313059b185
04_ebpf_kernel
[ "04_ebpf_kernel" ]
149
full
[ "kernelhook_max_one_program_attached", "ebpfprogram_attaches_kernelhook_xref", "ebpfprogram_subclass", "ebpfprogram_only_attached_to_kprobe" ]
[ 149, 38, 132 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,614
7,101
10,227
# Chapter 4: Kernel Hook and Program Attachment Constraints This chapter establishes the formal governance framework for managing eBPF kernel artifacts, specifically focusing on the axiomatic constraints governing the attachment of programs to kernel hooks. The scope encompasses the definition of artifact capacity, th...
1. **Analyze the Request:** * **Goal:** Write a single chapter of a technical textbook. * **Topic:** A small number of related concepts based on provided OWL ontology axioms. * **Style:** Professional technical documentation (audit reports, compliance handbooks, governance frameworks). Dense, eviden...
78c9af80-6d60-4f5a-b631-eddf6dac175a
9,437
1,848
5,080
2,779
0.01784
2026-06-08T02:22:45.737000
232ea5460ce6e0bf
4df17e1aae60b3f7
03_directive_governance
[ "03_directive_governance" ]
191
full
["attestation_with_supporting_evidence","control_with_evidence_requirement","requirement_equiv_speci(...TRUNCATED)
[ 191, 70, 126 ]
cerebras/zai-glm-4.7
glm
0.4
20,260,615
7,956
13,213
"# Chapter 4: Governance Directives and Evidence Management Frameworks\n\n## 4.1 Scope and Preamble\(...TRUNCATED)
"1. **Analyze the Request:**\n * **Goal:** Write a single chapter of a technical textbook.\n (...TRUNCATED)
c324cb25-cf2b-4a13-b6f7-ed2af32c7eaa
9,673
2,037
5,406
2,521
0.019162
2026-06-08T02:23:09.946000
232ea5460ce6e0bf
c858560fb308db74
03_directive_governance
[ "03_directive_governance", "07_long_tail" ]
5
full
["attestation_signed_by","gdpr_article_subclass","attestation_with_supporting_evidence","gdpr_data_s(...TRUNCATED)
[ 5, 151, 157 ]
xai/grok-4.3
grok
0.4
20,260,616
9,421
7,895
"# Chapter 4: Directive Governance and Attestation Structures in GDPR Compliance\n\nThis chapter def(...TRUNCATED)
"The task is to write a chapter of a technical reference book that teaches ontological concepts thro(...TRUNCATED)
07184f51-7c4a-4502-84bd-1e08d85be709
15,861
2,306
2,402
506
0.008753
2026-06-08T02:23:40.402000
232ea5460ce6e0bf
aadf6114f915f648
07_long_tail
[ "03_directive_governance", "07_long_tail" ]
35
full
["hipaa_safeguard_physical","control_only_addresses_threat_class","hipaa_safeguard_technical","hipaa(...TRUNCATED)
[ 35, 68, 127 ]
xai/grok-4.3
grok
0.4
20,260,617
9,654
6,245
"# Chapter 4: DirectiveICE Subclasses for HIPAA and PCI DSS Controls\n\nThis chapter defines five te(...TRUNCATED)
"The task is to write a chapter of a technical reference book that teaches ontological concepts thro(...TRUNCATED)
2e6e53b2-26a8-4ebd-b8da-5fa143f25000
15,497
2,336
2,484
1,044
0.008861
2026-06-08T02:24:10.309000
232ea5460ce6e0bf
End of preview. Expand in Data Studio

SDG Ontology-Grounded Synthetic Corpus — v0.3 (first cut)

A verifiable, attribution-clean synthetic textbook corpus for relational data-governance metadata, grounded in a BFO 2020 / CCO ontology. Unlike web-retrieval synthetic corpora, every chapter is generated from deterministic ontology axioms (not scraped text), populates a deterministic relational schema, and ships with the generator's reasoning trace.

Part of the Aegir project; the source artifacts live in zndx/sdg-corpora.

Contents

config / path rows description
chapters 2,235 textbook chapters — markdown prose + the model's reasoning trace, grounded in cited ontology axioms
columns 16,516 a blind relational column-classification benchmark — columns carry values only (anonymized ids); classify each into the vocabulary
vocabulary 548 the SKOS column-type label space (BFO/CCO-anchored: 7 upper concepts + 540 template leaves)
ontology/ 540 templates the source ontology (7 families) + sdg-vocab.ttl

Headline numbers: 9,230 relational tables · 16,516 columns · 79,746 cells · 5.79M reasoning tokens, produced for $34.

Why this is different (scale + quality vs the prior release)

axis v0.1 (sdg-bertopic-correspondence) v0.3 (this)
units 1,405 compositions 2,235 chapters
relational tables 0 9,230
columns / cells 0 / 0 16,516 / 79,746
reasoning tokens 0 5.79M
controlled vocabulary none 548-concept SKOS
verifiability one proxy (R≈0.52) DDL-valid + 98% structural + flat collapse-trend

Quality assurance

  • No collapse. A deterministic collapse-trend monitor (corpus distinct-n, gzip ratio, TF-IDF inter-chapter nearest-neighbour cosine, structure) binned by generation order is flat across the entire run — the size-controlled mode-collapse signal (intra-bin NN-cosine) holds at 0.5 end-to-end. 0 near-duplicates, 2 exact duplicates, 98% valid structured-table emission.
  • Verifiable. Every column is slot-typed from the ontology; the relational schema validates under Trino ∩ Spark. Claims trace to cited Manchester-syntax axioms.

The column-classification benchmark (columns)

A blind task: each column carries sampled values plus opaque table_id/column_idno table names, template ids, or labels. The label space is the vocabulary config (548 SKOS codes). The per-column ground-truth (column → SKOS code) is held out as the benchmark answer key. Predict a code per column from its values; request scoring against the held-out key.

Reasoning traces (for RL / distillation)

Every chapter in chapters includes response_reasoning — the generator's chain-of-thought (GLM-4.7 / Grok-4.3). 5.79M reasoning tokens total, paired with verifiable, ontology-grounded outputs.

Methodology

Ontology retrieval (verbalized Manchester axioms — the WHAT) + FinePDFs style anchors (the HOW) → weighted GLM-4.7 / Grok-4.3 mix → markdown chapter + populated DDL tables + a verifiable JSON schema → multi-scorer verification. Architecturally a descendant of textbook_quality (Vik Paruchuri), with web retrieval replaced by deterministic ontology retrieval so every claim is verifier-checkable.

License

Apache-2.0. The corpus is synthetic and attribution-clean (no scraped source text).

Downloads last month
53