# Auditing MCQA Benchmarks through Probability Landscapes

arXiv: https://arxiv.org/abs/2608.30372 · HTML: https://mlsanity.com/paper/2608.30372/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 30 references that are in the map.
- First posted on arXiv 31 Aug 2026.

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

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark](https://mlsanity.com/paper/2406.01574/) (2024)
- [CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge](https://mlsanity.com/paper/1811.00937/) (2018)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [Annotation Artifacts in Natural Language Inference Data](https://mlsanity.com/paper/1803.02324/) (2018)
- [Large Language Models Are Not Robust Multiple Choice Selectors](https://mlsanity.com/paper/2309.03882/) (2023)
- [Hypothesis Only Baselines in Natural Language Inference](https://mlsanity.com/paper/1805.01042/) (2018)
- [Generalization v.s. Memorization: Tracing Language Models' Capabilities Back to Pretraining Data](https://mlsanity.com/paper/2407.14985/) (2024)
- [Artifacts or Abduction: How Do LLMs Answer Multiple-Choice Questions Without the Question?](https://mlsanity.com/paper/2402.12483/) (2024)
- [When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation](https://mlsanity.com/paper/2602.16763/) (2026)
- [A BERT-based Distractor Generation Scheme with Multi-tasking and Negative Answer Training Strategies](https://mlsanity.com/paper/2010.05384/) (2020)
- [The Vulnerability of Language Model Benchmarks: Do They Accurately Reflect True LLM Performance?](https://mlsanity.com/paper/2412.03597/) (2024)
- [Better Distractions: Transformer-based Distractor Generation and Multiple Choice Question Filtering](https://mlsanity.com/paper/2010.09598/) (2020)
- [Rethinking Generative Large Language Model Evaluation for Semantic Comprehension](https://mlsanity.com/paper/2403.07872/) (2024)
- [BERT-based distractor generation for Swedish reading comprehension questions using a small-scale dataset](https://mlsanity.com/paper/2108.03973/) (2021)

## Built on by (most cited first, 1 of 1)

- [Are Benchmarks Reliable? Toward Structural Diagnosis via Item-Level Capability Boundaries](https://mlsanity.com/paper/2609.33196/) (2026)

## Abstract (arXiv)

As Large Language Models rapidly advance, performance on standard multiple-choice question answering (MCQA) benchmarks is reaching saturation. While the community has responded by developing increasingly difficult datasets, validating question quality and filtering flawed items remains a labor-intensive process. To provide a scalable diagnostic approach, we propose a two-component probabilistic framework for auditing MCQA benchmarks using model output distributions. First, for benchmark-level analysis, we characterize the probability landscape using the top prediction probability (Ptop1) and normalized residual entropy (Hnorm), summarized globally by Mean Pairwise Distance (MPD). Second, for item-level diagnostics, we introduce noise injection to reduce meaningful distractor competition, enabling us to flag candidate items for targeted human review and categorize residual failure patterns. Across four MCQA benchmarks, our landscape analysis reveals benchmark-level differences in model confidence and residual option competition. Concurrently, our noise-injection method flags potentially actionable item-level issues, showing alignment with expert error annotations from MMLU-Redux. These results suggest that our probability-based framework provides a lightweight audit lens for comparing macro-level benchmark structure and prioritizing individual items for targeted human review.