Measuring all the noises of LLM Evals

Why it matters (2 min)

• Separating signal from noise is central to experimental science.
• Applying well-established statistical method effectively to LLM evals requires consideration of their unique noise characteristics.
• Open receipts to verify and go deeper.

Go deeper (8 min)

Context

Separating signal from noise is central to experimental science. Applying well-established statistical method effectively to LLM evals requires consideration of their unique noise characteristics. We clearly define and measure three types of noise: prediction noise from generating different answers on a given question, data noise from sampling questions, and their combined total noise following the law of total variance. To emphasize relative comparisons and gain statistical power, we propose the all-pairs paired method, which applies the paired analysis to all pairs of LLMs and measures all the noise components based on millions of question-level predictions across many evals and settings. These measurements revealed clear patterns. First, each eval exhibits a characteristic and highly predictable total noise level across all model pairs. Second, paired prediction noise typically exceeds paired data noise, which means reducing prediction noise by averaging can significantly increase statistical power. These findings enable practitioners to assess significance without custom testing and to detect much smaller effects in controlled experiments.

For builders

Builder: scan the abstract + experiments; look for code, datasets, and evals.

Verify

Prefer primary announcements, papers, repos, and changelogs over reposts.

Receipts

1. Measuring all the noises of LLM Evals (arXiv stat.ML)