# Risk Is Not Review Value: Wrong-Answer Exposure Under Bounded Review Budgets

arXiv: https://arxiv.org/abs/2609.07095 · HTML: https://mlsanity.com/paper/2609.07095/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 33 references that are in the map.
- First posted on arXiv 7 Sep 2026.

## Builds on (most cited first, 20 of 20)

- [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)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [FEVER: a large-scale dataset for Fact Extraction and VERification](https://mlsanity.com/paper/1803.05355/) (2018)
- [FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation](https://mlsanity.com/paper/2305.14251/) (2023)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing](https://mlsanity.com/paper/2305.11738/) (2023)
- [Enabling Large Language Models to Generate Text with Citations](https://mlsanity.com/paper/2305.14627/) (2023)
- [How Can We Know When Language Models Know? On the Calibration of Language Models for Question Answering](https://mlsanity.com/paper/2012.00955/) (2020)
- [TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance](https://mlsanity.com/paper/2105.07624/) (2021)
- [SelectiveNet: A Deep Neural Network with an Integrated Reject Option](https://mlsanity.com/paper/1901.09192/) (2019)
- [Consistent Estimators for Learning to Defer to an Expert](https://mlsanity.com/paper/2006.01862/) (2020)
- [Measuring Attribution in Natural Language Generation Models](https://mlsanity.com/paper/2112.12870/) (2021)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [Predict Responsibly: Improving Fairness and Accuracy by Learning to Defer](https://mlsanity.com/paper/1711.06664/) (2017)
- [Fact-Checking the Output of Large Language Models via Token-Level Uncertainty Quantification](https://mlsanity.com/paper/2403.04696/) (2024)
- [Learning to Complement Humans](https://mlsanity.com/paper/2005.00582/) (2020)
- [Differentiable Learning Under Triage](https://mlsanity.com/paper/2103.08902/) (2021)
- [Calibrated Learning to Defer with One-vs-All Classifiers](https://mlsanity.com/paper/2202.03673/) (2022)
- [Classification Under Human Assistance](https://mlsanity.com/paper/2006.11845/) (2020)

## Abstract (arXiv)

LLM assistants often produce more answers than humans can review before users see them. Most evaluations ask whether an answer is wrong, unsupported, or low-confidence. Bounded review budgets instead ask which answers should be checked first under a fixed review budget. Risk alone is not enough: a high-risk answer may be hard to repair, while a moderately risky answer may be directly correctable from available evidence. For generated-answer evaluation, we model review prioritization as exposure reduction, where review value combines estimated wrongness, intervention affordance, impact, and cost. We evaluate review queues with Wrong-Answer Exposure Ratio (WAER), the fraction of wrong answers left unreviewed, and post-repair residual exposure (PRRE), the fraction still exposed after deterministic benchmark-supported repairs. PRRE uses repairability rules that do not numerically reuse the affordance scores used for ranking. On a 720-item TAT-QA/SciFact stress benchmark, review-value ranking keeps answer-level WAER nearly unchanged at 20% budget (0.605 vs. 0.600) but lowers PRRE from 0.881 to 0.716. These results show that trustworthy LLM evaluation should measure not only error detection, but also how limited review capacity reduces exposed wrong answers.