Geoffrey-Wang/

IntegriRef-Bench

Benchmark dataset for multi-level reference integrity verification, accompanying the IntegriRef framework.

	Dataset Splits




	

Split Rows Description

reference_verification 1,926 Golden benchmark + crawled verification cases (hallucinated, real, chimera, retracted)

signal_unit_tests 403 Per-signal unit tests for 14 Bayesian signal types

l1_intent 20 Citation intent classification test pairs

l2_nli 20 Claim-evidence NLI alignment test pairs… See the full description on the dataset page: https://huggingface.co/datasets/Geoffrey-Wang/IntegriRef-Bench.