# Cherry-pick Override: LLM Judges Under-use the Non-Directional Verdicts Their Contract Authorizes

arXiv: https://arxiv.org/abs/2606.07834 · HTML: https://mlsanity.com/paper/2606.07834/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 21 references that are in the map.
- First posted on arXiv 5 Jun 2026.

## Builds on (most cited first, 12 of 12)

- [Selective Classification for Deep Neural Networks](https://mlsanity.com/paper/1705.08500/) (2017)
- [SelectiveNet: A Deep Neural Network with an Integrated Reject Option](https://mlsanity.com/paper/1901.09192/) (2019)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [Get Your Vitamin C! Robust Fact Verification with Contrastive Evidence](https://mlsanity.com/paper/2103.08541/) (2021)
- [AVeriTeC: A Dataset for Real-world Claim Verification with Evidence from the Web](https://mlsanity.com/paper/2305.13117/) (2023)
- [Trust or Escalate: LLM Judges with Provable Guarantees for Human Agreement](https://mlsanity.com/paper/2407.18370/) (2024)
- [Tug-of-War Between Knowledge: Exploring and Resolving Knowledge Conflicts in Retrieval-Augmented Language Models](https://mlsanity.com/paper/2402.14409/) (2024)
- [The Automated Verification of Textual Claims (AVeriTeC) Shared Task](https://mlsanity.com/paper/2410.23850/) (2024)
- [ECon: On the Detection and Resolution of Evidence Conflicts](https://mlsanity.com/paper/2410.04068/) (2024)
- [Consensus or Conflict? Fine-Grained Evaluation of Conflicting Answers in Question-Answering](https://mlsanity.com/paper/2508.12355/) (2025)
- [Explaining Sources of Uncertainty in Automated Fact-Checking](https://mlsanity.com/paper/2505.17855/) (2025)
- [When Evidence Conflicts: Uncertainty and Order Effects in Retrieval-Augmented Biomedical Question Answering](https://mlsanity.com/paper/2605.14115/) (2026)

## Abstract (arXiv)

Evidence-grounded fact verification increasingly uses LLM judges to turn evidence into terminal verdicts. Many task contracts deliberately include non-directional verdicts - Conflicting for materially mixed evidence, Not Enough Evidence for absent evidence - so that a system can decline to assert a direction. Returning Supports or Refutes on a claim whose authorized verdict is non-directional is directional overcommitment: a label error whose output is a direction the contract did not authorize. Its conflict instance - both strands present, one of them returned - is Cherry-pick Override (CPO); unlike the insufficiency instance it cannot be repaired by retrieving more, because nothing was missing. On AVeriTeC's gold-Conflicting subset (N_C=150) a four-option typed panel commits directionally on 18.7% of claims, and four frozen contemporary judges (two proprietary, two open-weight) on 21.3-48.0%, with 24.0-38.0% on VitaminC-Mixed, a constructed stress test on which the ranking reverses. These are conservative lower bounds: the shared prompt specifies Conflicting in detail, forbids collapsing to a direction, and demonstrates the label three times. The failure is not specific to conflict: overcommitment on gold-Insufficient cases runs at a comparable rate for every judge and exceeds CPO for three of four on VitaminC-Mixed. Asked in a fresh context, two judges identify the conflict on about two thirds of their own CPO claims, against under 0.12 where they committed correctly: the recognition is available and does not reach the terminal action. Typed vocabulary, panel voting, confidence thresholds, a structural validator and self-decomposition each leave a substantial residual, and the last buys part of its reduction by relabelling one-sided evidence as conflicting. Every rate is dataset-defined; we report no human-confirmed prevalence.