What it measures

Large-scale benchmark for detecting and quantifying hallucinations across QA, dialogue, and summarization.

Reading the measurement

F1 balances precision and recall. Check both: a system can miss relevant evidence or include unsupported material, and those errors have different consequences. Keep the matching rule, averaging policy and denominator fixed.

Published evaluation

Macro hallucination rate (MaHR) is the share of responses containing a hallucination; micro hallucination rate (MiHR) averages hallucinated-fact proportions within responses. Both are lower-is-better and remain separate by domain; they are not hallucination-detection accuracy on original HaluEval.

Where this fits in an evaluation suite

Read accuracy together with error and abstention rates. Domain averages can hide language gaps, specialist failures, and poorly calibrated confidence.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Macro hallucination rate (MaHR) is the share of responses containing a hallucination; micro hallucination rate (MiHR) averages hallucinated-fact proportions within responses. Both are lower-is-better and remain separate by domain; they are not hallucination-detection accuracy on original HaluEval.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. The Dawn After the Dark: An Empirical Study on Factuality Hallucination in Large Language Models — published result table ↗

    Li, Junyi, Chen, Jie, Ren, Ruiyang, Cheng, Xiaoxue, Zhao, Wayne Xin, Nie, Jian-Yun et al. · 2024

    All 7 authors

    Li, Junyi, Chen, Jie, Ren, Ruiyang, Cheng, Xiaoxue, Zhao, Wayne Xin, Nie, Jian-Yun, Wen, Ji-Rong

Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.