Video Ad Script Constraint Validator Stress-Test Benchmark

#constraint verification #adversarial robustness evaluation #compliance classification #video advertising compliance #ad script quality assurance #advertising content validation

5000
0.5GB
JSON
CATL
MOBIUSI INC

Updated:2026-10-19

AI Analysis & Value Prop

This benchmark contains paired compliant and near-miss synthetic video ad scripts, covering boundary word counts, omitted required phrases, concealed prohibited terms, and attempts to evade surface-level checks. It provides explicit constraint rules, itemized validation results, and reward outcomes to assess rule following in controlled advertising text generation. The benchmark supports research on reinforcement learning with verifiable rewards (RLVR) and helps identify failure modes in which models exploit gaps in validator rules.

Dataset Insights

Sample Examples

{}

Technical Specifications

Field	Type	Description
file_name	string	File name
script_pair	object	A pair of original synthetic ad script texts, including a compliant example and a near-miss example, without encoding their verdicts in the field name.
reward_results	array	Reward outcomes for each script, calculated from the itemized validation results and reward rules.
failure_mode_label	string	A primary failure mode or stress-test label derived from the validation process to summarize validator weaknesses.
rule_specification	object	Original test rules and boundary conditions used to validate the scripts.
validation_results	array	Itemized verdicts and summary conclusions derived by applying the rules to each script in the pair.

Compliance Statement

Authorization Type	Proprietary - Commercial AI Training License (No Redistribution)
Commercial Use	Requires exclusive subscription or authorization contract (monthly or per-invocation charging)
Privacy and Anonymization	No PII, no real company names, simulated scenarios follow industry standards
Compliance System	Compliant with China's Data Security Law / EU GDPR / supports enterprise data access logs

Frequently Asked Questions

What does the dataset contain?: The dataset contains pairs of synthetic video ad scripts along with their constraint rules, itemized validation results, and reward outcomes. Test cases cover boundary word counts, omitted required phrases, concealed prohibited terms, and attempts to evade surface-level checks.

What evaluations is the dataset suitable for?: It can be used to evaluate whether controlled text generation systems and RLVR reward validators correctly check ad script constraints. It also supports analysis of boundary handling, validator gaps, and model failure modes that exploit checking rules.

Are the scripts real ads used in campaigns?: No. The scripts are synthetic text intended for controlled validator testing and research. They should not be treated as real advertising content or as data about campaign performance.

How can I access the data, and what are the license terms?: Refer to the dataset listing or provider for access methods, available formats, and license terms. Confirm that the license covers your intended research, evaluation, or commercial use before using the data.

Can't find the data you need?

Post a request and let data providers reach out to you.

Get this Dataset

Verified for Enterprise Use

Cite this Work

@dataset{Mobiusi2026,
  title={Video Ad Script Constraint Validator Stress-Test Benchmark},
  author={MOBIUSI INC},
  year={2026},
  url={https://www.mobiusi.com/datasets/a9f42d5b1f19ed54c91f32dbeceea924?dataset_scene_id=17},
  urldate={2026-10-12},
  keywords={video ad scripts, constraint validators, stress-test benchmark, synthetic text data, controlled text generation, RLVR, reward verification, rule-following evaluation},
  version={1.0}
}

Using this in research? Please cite us.

placeholder

Products

Scene

Domain

Modality

Task