# The Judge Knows When It Knows: Calibrated Abstention for LLM-Based A/B-Test Prediction

arXiv: https://arxiv.org/abs/2608.07517 · HTML: https://mlsanity.com/paper/2608.07517/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 21 references that are in the map.
- First posted on arXiv 2 Jul 2026.

## Builds on (most cited first, 9 of 9)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Generative Agents: Interactive Simulacra of Human Behavior](https://mlsanity.com/paper/2304.03442/) (2023)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [Large Language Models are not Fair Evaluators](https://mlsanity.com/paper/2305.17926/) (2023)
- [Selective Classification for Deep Neural Networks](https://mlsanity.com/paper/1705.08500/) (2017)
- [Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models](https://mlsanity.com/paper/2404.18796/) (2024)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [UXAgent: An LLM Agent-Based Usability Testing Framework for Web Design](https://mlsanity.com/paper/2502.12561/) (2025)

## Abstract (arXiv)

Can a multimodal LLM predict which version of a web page will win a real A/B test from screenshots alone? We report the most complete answer we are aware of, from six weeks of pre-registered experiments on real conversion tests: mostly no -- and the exceptions are identifiable in advance. On 330 real A/B tests a Gemini 3 Flash judge reaches Cohen's kappa = 0.14, but on the trustworthy (statistically significant) half of the labels the evidence is inconclusive (kappa = 0.11, CI includes zero). We show that 44% of the "ground-truth" labels in a leading CRO agency's catalog come from non-significant tests, and that the judge agrees more with the unreliable labels than the reliable ones -- a shared prior between labeler and model, not prediction. Every standard improvement lever (a 2.8x more expensive frontier model, prompt redesign, stimulus fidelity, change-type priors) fails its pre-registered gate. The judge's confident calls are different: a vote-margin gate isolates a subset (49% coverage) reaching kappa = 0.31 on significant labels. We measure the mechanism directly -- judges differing in model or prompt agree with each other at kappa = 0.74-0.88 while agreeing with real outcomes at only ~0.2, so a 16-vote panel carries about 2 effective independent votes -- and we reproduce it in humans: 15 CRO experts agree with each other (inter-rater kappa = 0.53) but score at chance against real outcomes (kappa ~ 0). Consensus, human or model, is reproducible, persuasive, and not evidence. We release our pre-registrations, locked gates, negative results, statistical harness, human responses, and a claims ledger in which every number carries an evidence tier.