cs.LGMay 25, 2026

Pre-Registering the Detectable Effect: A Paired-MDE Budget for 4-bit Quantization Benchmarks, with a Pilot Audit

Authors: Zexin Zhuang, Yanhang Li, Zhichao Fan

Organizations: 1Southern Methodist University · 2Northeastern University · University of Illinois Urbana-Champaign

Abstract

This is a planning-method note with an unpaired pilot audit. We adapt the classical paired-binary sample-size calculation (Miettinen, 1968) to quantization benchmarks, giving a conservative minimum detectable effect (MDE) bound δ∗≤(z1−α/2+z1−β)ρd/mδ^{*} \le (z_{1-α/2}+z_{1-β})\sqrt{ρ_d/m} in the paired item count mm and the FP16-NF4 disagreement rate ρdρ_d. The bound turns "how reliable is my quantization claim?" into a one-line budget a benchmark designer can commit to before running. We illustrate the bound on four models and four benchmarks (k=5k=5 splits of n=100n=100), and add a parallel MMLU prompt-template study to put the bound's quantization-noise scale alongside the prompt-noise scale. Assuming ρd=0.10ρ_d=0.10 (an unmeasured planning value), all observed NF4-FP16 deltas fall below the implied MDE, and most cross-split SDs lie within ±1.5\pm 1.5 pp of the binomial reference p(1−p)/n\sqrt{p(1-p)/n}, so much of the variance reported as "benchmark unreliability" on n=100n=100 subsamples is binomial sampling noise. The single borderline cell (OPT-WinoGrande, ∣Δ∣=3.2|Δ|=3.2 pp) is below the implied MDE at ρd=0.10ρ_d=0.10 but above it at ρd=0.05ρ_d=0.05, illustrating the planning trade-off the bound makes explicit. On MMLU, prompt-template ranges of 2-10 pp meet or exceed the largest observed quantization delta (3.2 pp), so a quantization audit that does not first fix the prompt template absorbs template variance into its noise floor. We complement the bound with a five-line pre-registration template.

Explore similar work

CardsList