Datasets:

leeroy-jankins
/

Inspector-General-Act-of-1978

pretty_name: Inspector General Act of 1978 Question Answering Dataset
language:
  - en
license: other
task_categories:
  - question-answering
  - text-generation
tags:
  - inspector-general-act
  - public-law-95-452
  - inspectors-general
  - federal-oversight
  - government-accountability
  - audits
  - investigations
  - fraud-prevention
  - waste-and-abuse
  - whistleblowers
  - federal-law
  - united-states-congress
  - legal-question-answering
  - document-question-answering
  - instruction-tuning
  - synthetic-data
size_categories:
  - n<1K

Dataset Description

The Inspector General Act of 1978 Question Answering Dataset is a document-grounded collection of 150 question-and-answer records concerning the original statutory framework used to establish independent Offices of Inspector General within selected executive-branch departments and agencies.

The dataset was developed from the enacted text of the Inspector General Act of 1978, Public Law 95-452, 92 Stat. 1101, approved on October 12, 1978. The statute established Offices of Inspector General, assigned audit and investigative responsibilities, created reporting requirements, granted access and subpoena authorities, protected employee complainants, transferred existing oversight organizations, and defined the relationship between Inspectors General, agency leadership, Congress, and other federal oversight institutions.

Each record contains a natural-language question and a detailed answer derived from the enacted legislation. The questions require substantive understanding of the law rather than identification of page numbers, section locations, headings, or isolated statutory phrases. Every answer contains at least five complete sentences and is designed to provide sufficient context for independent understanding.

The dataset may support supervised fine-tuning, retrieval-augmented generation, legal question answering, statutory-comprehension evaluation, federal-oversight research, semantic retrieval, instruction tuning, and testing of systems that answer questions from enacted public laws.

Dataset Summary

Attribute	Value
Domain	Federal oversight, auditing, investigations, and government accountability
Jurisdiction	United States
Legislative body	95th United States Congress
Source type	Enacted public law
Public law	Public Law 95-452
Bill number	H.R. 8588
Statutes at Large citation	92 Stat. 1101
Date approved	October 12, 1978
Statutory effective date	October 1, 1978
Language	English
Number of records	150
Primary task	Document-grounded question answering
Secondary task	Instruction-oriented text generation
Answer length	At least five complete sentences per record
Data type	Structured textual question-and-answer pairs
Generation method	Human-directed synthetic dataset generation
External knowledge required	No

Source Material

The dataset is based on the enacted version of:

United States Congress. Inspector General Act of 1978. Public Law 95-452, 92 Stat. 1101. Approved October 12, 1978.

The Act’s stated purpose was to reorganize portions of the executive branch and increase governmental economy and efficiency by establishing Offices of Inspector General within designated departments, agencies, and administrations.

The dataset is grounded in the original 1978 enactment. It does not incorporate every subsequent amendment, recodification, expansion, or organizational change affecting the federal Inspector General system.

Users seeking to apply the law to current circumstances should consult the current codified provisions, later amendments, applicable agency-specific statutes, and authoritative legal guidance.

Original Establishments Covered

The original Act established Offices of Inspector General within the following departments:

Department of Agriculture.
Department of Commerce.
Department of Housing and Urban Development.
Department of the Interior.
Department of Labor.
Department of Transportation.

It also established offices within the following agencies and administrations:

Community Services Administration.
Environmental Protection Agency.
General Services Administration.
National Aeronautics and Space Administration.
Small Business Administration.
Veterans’ Administration.

The composition, names, and legal status of some of these organizations subsequently changed. The dataset reflects the terminology used in the original enactment.

Dataset Scope

The records address the major substantive areas of the original Act.

Purpose and Establishment

Records may address:

Creation of independent and objective oversight units.
Audits and investigations of federal programs and operations.
Promotion of economy, efficiency, and effectiveness.
Prevention and detection of fraud and abuse.
Reporting to agency heads and Congress.
Corrective-action recommendations.
Establishment of Offices of Inspector General.
Original departments and agencies covered by the statute.

Appointment and Independence

Records may address:

Presidential appointment.
Senate advice and consent.
Nonpartisan selection.
Professional qualifications.
Integrity requirements.
Reporting relationships.
General supervision by establishment heads.
Restrictions on supervision by other agency officials.
Prohibition on blocking audits and investigations.
Subpoena independence.
Presidential removal.
Congressional notice of removal reasons.
Political-activity classifications.

Office Leadership

Records may address:

Assistant Inspector General for Auditing.
Assistant Inspector General for Investigations.
Supervision of audit activities.
Supervision of investigative activities.
Civil-service requirements.
Separation of audit and investigative leadership.

Duties and Responsibilities

Records may address:

Policy direction for audits and investigations.
Conduct, supervision, and coordination of oversight work.
Review of existing legislation.
Review of proposed legislation.
Review of existing and proposed regulations.
Recommendations concerning economy and efficiency.
Recommendations concerning fraud prevention.
Coordination with federal agencies.
Coordination with state and local governments.
Coordination with nongovernmental organizations.
Identification and prosecution of participants in fraud.
Corrective-action monitoring.
Reporting significant problems and deficiencies.

Audit Standards and Coordination

Records may address:

Comptroller General audit standards.
Use of nonfederal auditors.
Quality requirements for contracted audit work.
Avoidance of duplicate oversight.
Cooperation with the Comptroller General.
Coordination among federal oversight institutions.

Criminal Referrals

Records may address:

Reasonable grounds to suspect a federal criminal violation.
Expeditious reporting to the Attorney General.
Relationship between Inspector General investigations and prosecutive authorities.
Distinction between referral and prosecution.

Semiannual Reports

Records may address:

April 30 reporting deadline.
October 31 reporting deadline.
Six-month reporting periods.
Significant problems, abuses, and deficiencies.
Corrective recommendations.
Unresolved prior recommendations.
Matters referred for prosecution.
Resulting prosecutions and convictions.
Reports concerning refused information.
Lists of completed audit reports.
Transmission by establishment heads.
Agency-head comments.
Thirty-day congressional transmission requirement.
Sixty-day public-availability requirement.

Immediate Reporting

Records may address:

Particularly serious or flagrant problems.
Immediate reporting to the establishment head.
Seven-calendar-day congressional transmission.
Agency comments accompanying urgent reports.
Differences between immediate and semiannual reports.

Statutory Authorities

Records may address:

Access to agency records.
Authority to conduct investigations.
Authority to prepare reports.
Requests for intergovernmental assistance.
Subpoena authority.
Federal district-court enforcement.
Restrictions on using subpoenas against federal agencies.
Direct and prompt access to establishment heads.
Personnel appointment authority.
Expert and consulting services.
Contracting for audits, studies, and analyses.
Advance appropriations requirements.

Agency Cooperation

Records may address:

Duties of federal agencies receiving information requests.
Existing statutory or regulatory restrictions.
Unreasonable refusal of information.
Reporting noncooperation to the establishment head.
Office space and facility requirements.
Central and field-office support.
Equipment, supplies, communications, and maintenance.

Employee Complaints

Records may address:

Violations of law, rules, or regulations.
Mismanagement.
Gross waste of funds.
Abuse of authority.
Public-health and safety dangers.
Confidentiality of complainant identities.
Unavoidable disclosure during investigations.
Prohibition on reprisal.
Threatened personnel actions.
Knowingly false complaints.
Willful disregard for truth or falsity.

Department of Defense Transitional Provisions

Records may address:

Temporary Defense Department semiannual reports.
Reporting periods and deadlines.
Fraud, waste, and abuse reporting.
Prosecution summaries.
Audit and inspection statistics.
Public availability.
National-security exclusions.
Intelligence-function exclusions.
Notice to congressional committee leadership.
Delegation to Senate-confirmed presidential appointees.
Proposed post-1982 reporting legislation.
Establishment of the Defense task force.
Task-force independence.
Task-force access to information.
Protection of active investigations.
Removal of names from reports.
Security-clearance requirements.
Review of oversight-component independence.
Relationship with the Department of Justice.
Organizational recommendations.
Submission of the final task-force report.
Disestablishment of the task force.
Exclusion of certain intelligence and counterintelligence matters.

Transfer of Functions

Records may address:

Transfer of existing audit offices.
Transfer of existing investigative offices.
Transfer of inspection organizations.
Agency-specific transfers.
Additional transfers authorized by establishment heads.
Prohibition on transferring program operating responsibilities.
Transfer of personnel.
Transfer of records and property.
Transfer of contracts and liabilities.
Transfer of unexpended funds.
Employee classification and compensation protection.
Lapse of predecessor offices.
Continuing-pay protections.

Conforming and Technical Amendments

Records may address:

Addition of Inspector General positions to executive pay statutes.
Addition of a Deputy Inspector General position.
Technical correction of statutory cross-references.
Integration of new offices into existing federal personnel law.

Definitions and Effective Date

Records may address:

Head of the establishment.
Establishment.
Inspector General.
Office.
Federal agency.
Exclusion of the General Accounting Office.
Statutory effective date.
Relationship between enactment and effective dates.

Dataset Structure

Each row represents one question-and-answer exchange.

Data Fields

Field	Type	Description
`ID`	integer	Sequential record identifier ranging from 1 through 150.
`U`	string	Role label for the question author. The value is `User`.
`Question`	string	A natural-language question answerable from the original Act.
`A`	string	Role label for the responding system. The value is `Agent`.
`Answer`	string	A document-grounded answer containing at least five complete sentences.

Example Record

{
  "ID": 1,
  "U": "User",
  "Question": "What broad governmental problem was the Inspector General Act designed to address?",
  "A": "Agent",
  "Answer": "The Act was designed to improve economy and efficiency within the executive branch. It created independent and objective units to audit and investigate federal programs and operations. It also sought to prevent and detect fraud and abuse. Another objective was to ensure that agency leaders and Congress received current information about serious problems and deficiencies. The statute therefore combined oversight, accountability, and corrective-action functions within a permanent institutional structure."
}

Dataset Creation

Generation Process

The dataset was created through the following process:

The original enacted public law was reviewed by section and substantive requirement.
The Act’s institutional structure, appointment provisions, reporting duties, statutory authorities, complaint protections, transitional arrangements, transfer provisions, and definitions were identified.
Questions were drafted to test substantive understanding of the enacted provisions.
Questions limited to locating a page, section, paragraph, heading, or physical portion of the source were avoided.
Answers were generated using information contained in the original Act.
Each answer was written as a self-contained explanation containing at least five complete sentences.
Records were reviewed for statutory grounding, topical coverage, relevance, internal consistency, and duplication.
The final records were assigned unique sequential identifiers from 1 through 150.

Nature of the Data

The records are synthetic instructional examples. They are not transcripts of Inspectors General, auditors, investigators, whistleblowers, agency employees, members of Congress, prosecutors, or private parties.

The dataset does not contain:

Personally identifiable information.
Actual employee complaints.
Confidential whistleblower information.
Law-enforcement-sensitive records.
Classified information.
Active investigative material.
Actual subpoena returns.
Grand-jury information.
Agency deliberative material.
Personnel records.
Nonpublic audit working papers.
Information about identifiable subjects of investigation.

Quality Controls

The dataset was designed using the following controls:

Every question is intended to be answerable from the original enacted legislation.
Every answer contains at least five complete sentences.
Answers use complete grammatical prose.
Questions avoid document-location phrasing.
Questions emphasize statutory meaning, requirements, protections, and institutional relationships.
Records cover all major sections of the source.
Questions were drafted to minimize semantic duplication.
Answers avoid knowingly introducing later amendments as though they appeared in the original law.
Role labels remain consistent across all records.
Record identifiers are sequential and unique.

Supported Tasks

Document-Grounded Question Answering

The dataset may support systems that answer questions using the Inspector General Act of 1978 as the authoritative context.

Potential applications include:

Closed-book question answering after supervised fine-tuning.
Open-book question answering using retrieved statutory passages.
Retrieval-augmented generation evaluation.
Legal and legislative research assistant prototyping.
Federal-oversight knowledge-base development.
Statutory-comprehension testing.
Answer-grounding evaluation.
Hallucination detection.
Public-administration education.
Government-accountability research.

Instruction Tuning

The records may be transformed into a conversational format.

{
  "messages": [
    {
      "role": "user",
      "content": "What limits does the Act place on agency supervision of an Inspector General?"
    },
    {
      "role": "assistant",
      "content": "The establishment head may exercise general supervision over the Inspector General. However, that official may not prevent or prohibit the Inspector General from initiating an audit or investigation. The same protection applies to carrying out or completing the work. The agency head also may not prevent the issuance of a subpoena during an audit or investigation. These restrictions protect the operational independence of the office."
    }
  ]
}

Retrieval-Augmented Generation

The questions may be used as retrieval queries against segmented portions of the public law.

The dataset may be useful for evaluating whether a retrieval system distinguishes among:

General supervision and prohibited interference.
Appointment and removal.
Auditing and investigations.
Semiannual and immediate reports.
Agency-head comments and Inspector General findings.
Requests for information and subpoenas.
Federal-agency cooperation and judicial enforcement.
Employee confidentiality and reprisal protection.
Agency oversight and program operating responsibilities.
General Inspector General provisions and temporary Defense provisions.
Original enactment and later federal Inspector General law.

Legal Text Generation

The answers provide examples of extended explanations grounded in enacted statutory language. They may be used to evaluate whether a model can explain legal provisions without introducing later amendments, unsupported procedural requirements, or policy conclusions not contained in the source.

Semantic Search and Reranking

The questions may be paired with statutory passages to create:

Dense-retrieval tasks.
Sparse-retrieval tasks.
Hybrid-retrieval tasks.
Passage-reranking tasks.
Multi-section answer-synthesis tasks.
Citation-selection evaluations.

Loading the Dataset

Replace the placeholder with the final Hugging Face organization or username and repository name.

from datasets import load_dataset

dataset = load_dataset(
    "<organization-or-username>/<dataset-repository>"
)

print(dataset)
print(dataset["train"][0])

To load a CSV file stored in the repository:

from datasets import load_dataset

dataset = load_dataset(
    "csv",
    data_files="data.csv"
)

print(dataset["train"][0])

To normalize the original column names:

from datasets import load_dataset

dataset = load_dataset(
    "<organization-or-username>/<dataset-repository>"
)

dataset = dataset.rename_columns(
    {
        "ID": "id",
        "Question": "question",
        "Answer": "answer"
    }
)

dataset = dataset.remove_columns(
    [
        "U",
        "A"
    ]
)

print(dataset["train"].features)

To convert the records into a conversational format:

from datasets import load_dataset

dataset = load_dataset(
    "<organization-or-username>/<dataset-repository>",
    split="train"
)

def create_messages(record):
    return {
        "messages": [
            {
                "role": "user",
                "content": record["Question"]
            },
            {
                "role": "assistant",
                "content": record["Answer"]
            }
        ]
    }

chat_dataset = dataset.map(
    create_messages,
    remove_columns=dataset.column_names
)

print(chat_dataset[0])

Recommended Data Splits

Because the dataset contains 150 records, it may be published as a single train split.

For experimental evaluation, deterministic training, validation, and test subsets may be created:

from datasets import DatasetDict, load_dataset

dataset = load_dataset(
    "<organization-or-username>/<dataset-repository>",
    split="train"
)

train_test = dataset.train_test_split(
    test_size=0.20,
    seed=42
)

validation_test = train_test["test"].train_test_split(
    test_size=0.50,
    seed=42
)

splits = DatasetDict(
    {
        "train": train_test["train"],
        "validation": validation_test["train"],
        "test": validation_test["test"]
    }
)

print(splits)

A random split may place closely related statutory concepts in separate subsets. Topic-based splitting may provide a more demanding evaluation of generalization.

Potential topic groups include:

Purpose and establishment.
Appointment and independence.
Duties and responsibilities.
Audit standards.
Criminal referrals.
Semiannual reports.
Immediate reports.
Access and subpoena authority.
Employee complaints.
Department of Defense transitional provisions.
Transfer of functions.
Technical amendments.
Definitions and effective date.

Intended Uses

The dataset is intended for:

Research involving document-grounded language models.
Development of legal question-answering systems.
Federal-oversight and accountability research.
Retrieval-augmented generation experiments.
Semantic-search evaluation.
Statutory-comprehension benchmarks.
Instruction-tuning experiments.
Evaluation of source fidelity and answer completeness.
Prototyping public-law knowledge assistants.
Public-administration education.
Testing whether models distinguish closely related oversight concepts.
Evaluation of citation and passage-selection systems.

Out-of-Scope Uses

The dataset is not intended to:

Replace the official public law.
Represent the complete current Inspector General Act.
Provide legal advice.
Determine the present authority of a particular Inspector General.
Resolve an active subpoena dispute.
Determine whether information may lawfully be withheld.
Adjudicate an employee reprisal complaint.
Identify or investigate an actual whistleblower.
Determine whether a federal crime occurred.
Direct a criminal referral or prosecution.
Resolve questions involving classified or law-enforcement-sensitive information.
Determine the current reporting obligations of a specific agency without reviewing later law.
Make autonomous personnel, investigative, prosecutorial, or oversight decisions.
Substitute for qualified legal, audit, investigative, or congressional review.

Limitations

Original-Enactment Scope

The dataset is based on the Inspector General Act as enacted in 1978. The federal Inspector General framework has been amended, expanded, reorganized, and recodified since that time.

An answer may accurately describe the original Act without accurately describing current law.

Historical Agency Names

The source uses agency names and organizational structures existing in 1978. Some named entities were later renamed, reorganized, abolished, or succeeded by other organizations.

The dataset preserves the source terminology where appropriate.

Synthetic Questions and Answers

The records are constructed instructional examples. They do not reproduce the natural distribution, terminology, uncertainty, or urgency of questions asked by auditors, investigators, agency counsel, congressional staff, employees, or subjects of investigations.

Single-Source Emphasis

The dataset is grounded primarily in Public Law 95-452. Complete legal analysis may require later amendments, current codification, agency-specific Inspector General statutes, appropriations laws, case law, regulations, and administrative guidance.

No Independent Legal Interpretation

The answers summarize and explain the original legislation. They do not constitute official legal interpretations or opinions.

Answer-Length Style

Every answer contains at least five sentences because of the dataset’s construction requirements. Some answers may therefore be more detailed than necessary for practical use.

Limited Size

The dataset contains 150 records. It may support evaluation, demonstrations, retrieval experiments, augmentation, or narrow fine-tuning, but it is not sufficient by itself to train a broadly capable legal or federal-oversight model.

Potential Paraphrasing Errors

Paraphrasing can omit qualifications or create ambiguity. Consequential conclusions should be verified against the enacted text and current law.

Uneven Topic Density

The Act devotes more detail to certain subjects, including reports, authorities, Department of Defense transitional arrangements, and transfers of functions. The number of records devoted to each subject may therefore vary.

Bias and Risk Considerations

The dataset inherits the institutional structure, terminology, assumptions, and policy objectives of the original statute.

Potential risks include:

Treating the original 1978 framework as current law.
Ignoring later amendments or recodification.
Confusing an Inspector General’s investigative authority with prosecutorial authority.
Treating an allegation as proof of wrongdoing.
Assuming all employee complaints must be investigated.
Treating confidentiality of complainant identity as absolute.
Overlooking statutory or regulatory restrictions on agency information.
Confusing a request for records with a subpoena.
Confusing an Inspector General office with a program-management office.
Assuming an establishment head may direct the result of an audit.
Using model-generated answers to guide an active investigation.
Disclosing sensitive or identifying information in a public system.
Treating generalized explanations as legal determinations.

Models trained or evaluated with the dataset may generate fluent but incomplete legal explanations. Human review is required for consequential applications.

Privacy and Confidentiality Considerations

The dataset contains no actual employee complaints or identifiable investigative subjects.

Users should not combine the dataset with private or sensitive information in a way that:

Reveals a whistleblower’s identity.
Exposes an active investigation.
Discloses law-enforcement-sensitive information.
Reveals classified information.
Publishes protected personnel records.
Circumvents established complaint channels.
Interferes with criminal or administrative proceedings.

Applications should distinguish public statutory knowledge from protected case information and use appropriate access controls.

Responsible Use

Users should:

Verify important conclusions against the official public law.
Consult the current codified Inspector General provisions.
Review subsequent amendments and agency-specific authorities.
Distinguish the original 1978 Act from present law.
Avoid entering protected complaint or investigative information into unauthorized systems.
Maintain human oversight for legal, audit, investigative, and personnel matters.
Avoid treating model output as a criminal finding or prosecutorial decision.
Preserve conditions, exceptions, deadlines, and reporting relationships.
Evaluate systems for hallucinated authorities, deadlines, agencies, and legal standards.
Clearly distinguish educational summaries from official legal advice.

Evaluation Considerations

Groundedness

The answer should remain supported by the original enacted legislation. It should not introduce later amendments as though they appeared in Public Law 95-452.

Statutory Accuracy

The response should correctly identify:

The responsible official.
The applicable reporting deadline.
The recipient of a report.
The scope of access authority.
The conditions for subpoena use.
The limits on agency supervision.
The requirements for employee confidentiality.
The conditions governing reprisal protection.
The authorities transferred to each office.
The temporary nature of Defense provisions.

Completeness

The answer should include material conditions, exceptions, deadlines, and procedural relationships needed to explain the provision accurately.

Terminological Precision

The system should distinguish among:

Inspector General and Office of Inspector General.
Audit and investigation.
General supervision and operational interference.
Agency-head comments and Inspector General findings.
Semiannual and immediate reports.
Request for information and subpoena.
Referral and prosecution.
Complaint and substantiated violation.
Oversight responsibility and program operating responsibility.
Establishment head and Inspector General.
Original statutory provisions and current law.

Historical Precision

The system should preserve the original names of agencies and positions when explaining the 1978 Act. It should not silently substitute modern organizational names unless the distinction is clearly explained.

Citation Accuracy

When source passages are available, citations should point to the provisions that directly support the answer.

Multi-Section Reasoning

Some questions require combining appointment, independence, reporting, access, or transfer provisions. Systems should be evaluated on their ability to synthesize those provisions without inventing additional authority.

Hallucination Resistance

A model should acknowledge when a question requires information not provided in the original Act, such as:

Current agency organization.
Later statutory amendments.
Current reporting formats.
Current whistleblower-remedy procedures.
Case-specific investigative information.
Judicial interpretations.
Current appropriations.
Present executive-pay classifications.

Suggested Evaluation Metrics

Automated evaluation may include:

Exact match for narrowly framed factual questions.
Token-level F1.
ROUGE.
BERTScore.
Semantic similarity.
Retrieval recall at selected cutoff values.
Mean reciprocal rank.
Normalized discounted cumulative gain.
Citation precision.
Citation recall.
Natural-language-inference support scores.
LLM-assisted groundedness scoring.
Human legal-accuracy review.
Human completeness review.
Human historical-accuracy review.

Automated similarity metrics should not be used alone because multiple legally accurate answers may differ substantially in wording.

Recommended Retrieval Evaluation Format

An enriched retrieval version may include:

Field	Description
`question`	The retrieval query.
`answer`	The reference answer.
`section`	The applicable section of the Act.
`subsection`	The applicable subsection or paragraph.
`source_text`	The statutory passage supporting the answer.
`topic`	The substantive subject category.
`affected_establishment`	The department or agency implicated by the provision.
`reporting_deadline`	A statutory deadline, when applicable.
`requires_multiple_passages`	Whether the answer requires synthesis of several provisions.
`difficulty`	An optional qualitative difficulty label.

These fields are not included in the original five-column dataset and would require a separately enriched release.

Potential Topic Labels

Records may be labeled with values such as:

purpose
establishment
appointment
qualifications
independence
removal
audit_leadership
investigative_leadership
duties
legislative_review
audit_standards
criminal_referrals
semiannual_reports
immediate_reports
records_access
subpoena_authority
agency_cooperation
staffing
employee_complaints
confidentiality
reprisal
defense_reporting
defense_task_force
transfer_of_functions
pay_protection
technical_amendments
definitions
effective_date

Maintenance

The dataset should be reviewed when:

Errors are identified in the generated records.
The repository is revised to cover current law.
Later Inspector General amendments are added.
Current codified provisions are incorporated.
Agency names or historical notes require clarification.
Source passages and citations are added.
Topic labels or retrieval fields are introduced.

Recommended versioning practices include:

Preserving the original 150-record release.
Clearly labeling it as based on the original 1978 enactment.
Maintaining a change log.
Identifying corrected, removed, or replaced records.
Separating historical-source revisions from stylistic edits.
Publishing a distinct dataset version for current law.
Recording the source version used for every release.

Licensing Information

The underlying source is an enacted United States federal law published as Public Law 95-452.

The generated questions, answers, dataset organization, and documentation are derivative dataset content prepared from public federal source material.

The repository uses the metadata value license: other because users should independently evaluate the legal status of the underlying federal material, the generated compilation, and their intended use. No separate ownership claim is made over the underlying statutory text.

Users remain responsible for complying with applicable laws, repository terms, attribution practices, and downstream-use requirements.

Citation

When using the dataset, cite both the dataset repository and the underlying public law.

@dataset{inspector_general_act_1978_qa,
  title        = {Inspector General Act of 1978 Question Answering Dataset},
  year         = {2026},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/<organization-or-username>/<dataset-repository>},
  note         = {A 150-record document-grounded question-answering dataset derived from Public Law 95-452}
}

@techreport{inspector_general_act_1978,
  author       = {{United States Congress}},
  title        = {Inspector General Act of 1978},
  institution  = {United States Government Printing Office},
  number       = {Public Law 95-452},
  year         = {1978},
  month        = {October},
  note         = {92 Stat. 1101; approved October 12, 1978}
}

Suggested Legal Citation

Inspector General Act of 1978, Pub. L. No. 95-452, 92 Stat. 1101.

Acknowledgments

The underlying legislation was enacted by the 95th United States Congress and published as a federal public law.

The dataset is an independently prepared question-answering resource. It should not be interpreted as an official congressional, executive-branch, Inspector General, Department of Justice, Government Accountability Office, or agency dataset, endorsement, interpretation, audit product, investigative report, or legal opinion.

Disclaimer

This dataset is provided for research, education, evaluation, and technical experimentation.

It does not constitute legal advice, an official interpretation of federal law, a criminal finding, a subpoena, a prosecutorial referral, an audit opinion, an investigative determination, a whistleblower-protection decision, or authorization to take personnel or enforcement action.

The official public law, current codified statutes, subsequent amendments, applicable agency-specific authorities, judicial decisions, and authoritative legal guidance control whenever a difference exists between this dataset and an authoritative source.