Related papers: With Little Power Comes Great Responsibility

Large Language Models (LLMs) frequently hallucinate plausible but incorrect assertions, a vulnerability often missed by uncertainty metrics when models are confidently wrong. We propose DiffuTruth, an unsupervised framework that…

Underpowered studies (below 50% power) suffer from the winner's curse: A statistically significant positive estimate must exaggerate the true treatment effect to meet the significance threshold. A study by Dipayan Biswas, Annika Abell, and…

This paper considers the asymptotic power of likelihood ratio test (LRT) for the identity test when the dimension p is large compared to the sample size n. The asymptotic distribution of LRT under alternatives is given and an explicit…

Reliable certification of Large Language Models (LLMs)-verifying that failure rates are below a safety threshold-is critical yet challenging. While "LLM-as-a-Judge" offers scalability, judge imperfections, noise, and bias can invalidate…

A central question in natural language understanding (NLU) research is whether high performance demonstrates the models' strong reasoning capabilities. We present an extensive series of controlled experiments where pre-trained language…

We consider linear regression in the high-dimensional regime where the number of observations $n$ is smaller than the number of parameters $p$. A very successful approach in this setting uses $\ell_1$-penalized least squares (a.k.a. the…

Existing approaches to few-shot learning in NLP rely on large language models (LLMs) and/or fine-tuning of these to generalise on out-of-distribution data. In this work, we propose a novel few-shot learning approach based on soft-label…

Statistical significance testing centered on p-values is commonly used to compare NLP system performance, but p-values alone are insufficient because statistical significance differs from practical significance. The latter can be measured…

Refining one's hypotheses in the light of data is a common scientific practice; however, the dependency on the data introduces selection bias and can lead to specious statistical analysis. An approach for addressing this is via conditioning…

In multiple testing several criteria to control for type I errors exist. The false discovery rate, which evaluates the expected proportion of false discoveries among the rejected null hypotheses, has become the standard approach in this…

Large language models (LLMs) are widely used as zero-shot and few-shot classifiers, where task behaviour is largely controlled through prompting. A growing number of works have observed that LLMs are sensitive to prompt variations, with…

Recently, several authors have re-examined the power of the classical F-test in linear regression in a `large-p, large-n' framework (cf. Zhong and Chen (2011), Wang and Cui (2013)). They highlight the loss of power as the number of…

Robust tests of general composite hypothesis under non-identically distributed observations is always a challenge. Ghosh and Basu (2018, Statistica Sinica, 28, 1133--1155) have proposed a new class of test statistics for such problems based…

Multivariate linear regressions are widely used statistical tools in many applications to model the associations between multiple related responses and a set of predictors. To infer such associations, it is often of interest to test the…

In the last few years, the ML community has created a number of new NLP models based on transformer architecture. These models have shown great performance for various NLP tasks on benchmark datasets, often surpassing SOTA results. Buoyed…

Large language models (LLMs) have demonstrated significant capability to generalize across a large number of NLP tasks. For industry applications, it is imperative to assess the performance of the LLM on unlabeled production data from time…

We consider the hypothesis testing problem of detecting a shift between the means of two multivariate normal distributions in the high-dimensional setting, allowing for the data dimension p to exceed the sample size n. Specifically, we…

Statistical insignificance does not suggest the absence of effect, yet scientists must often use null results as evidence of negligible (near-zero) effect size to falsify scientific hypotheses. Doing so must assess a result's null strength,…

Standard NLP benchmarks often fail to capture vulnerabilities stemming from dataset artifacts and spurious correlations. Contrast sets address this gap by challenging models near decision boundaries but are traditionally labor-intensive to…

Researchers often misinterpret and misrepresent statistical outputs. This abuse has led to a large literature on modification or replacement of testing thresholds and $P$-values with confidence intervals, Bayes factors, and other devices.…