Datasetpublic

VeritasOmega/veritas-claim-verification-dataset

VERITAS Claim Verification Dataset 775,294 labeled verification calls (770,000 synthetic + 5,294 adversarial)Multi-model consensus verdicts from the VERITAS Oracle across 8 domains. Paper VERITAS: Independence-Weighted Multi-Model Consensus for AI Output Verification Adversarial Robustness Findings (NEW) 5,294 adversarial variants generated via deterministic transforms: Transform N Accuracy negation 2,379 99.7% jurisdiction 114 100.0%… See the full description on the dataset page: https://huggingface.co/datasets/VeritasOmega/veritas-claim-verification-dataset.

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face

0likes6downloads

Dataset Card

VERITAS Claim Verification Dataset

775,294 labeled verification calls (770,000 synthetic + 5,294 adversarial) Multi-model consensus verdicts from the VERITAS Oracle across 8 domains.

Paper

VERITAS: Independence-Weighted Multi-Model Consensus for AI Output Verification

Adversarial Robustness Findings (NEW)

5,294 adversarial variants generated via deterministic transforms:

Transform	N	Accuracy
negation	2,379	99.7%
jurisdiction	114	100.0%
quantifier	337	98.8%
number_perturb	85	98.8%
temporal	2,379	79.8%
Overall adversarial	5,294	90.7%
Baseline		98.4%
Delta		-7.7pp

VERITAS is highly robust to negation attacks (99.7%) — a category where single-model verifiers typically drop 10-15pp. Temporal hedging (79.8%) is the documented weakness: no time-indexed knowledge retrieval.

Datasets

—eritasoracle.syntheticdataset — 770k synthetic claims
—eritasoracle.syntheticdataset_adversarial — 5,294 adversarial variants with transform labels

Schema

claimhash, claimtext, domain, groundtruth, source, confidence, consensus, flags, modelvotes, traceid, latencyms, correct, ingested_at

Citation

@misc{lopez2026veritas, title={VERITAS: Independence-Weighted Multi-Model Consensus for AI Output Verification}, author={RJ Lopez}, year={2026}, doi={10.5281/zenodo.19187611} }