# Benchmarks Saturate When The Model Gets Smarter Than The Judge

arXiv: https://arxiv.org/abs/2601.19532 · HTML: https://mlsanity.com/paper/2601.19532/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 30 references that are in the map.
- First posted on arXiv 27 Jan 2026.

## Builds on (most cited first, 25 of 25)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [A Survey on Evaluation of Large Language Models](https://mlsanity.com/paper/2307.03109/) (2023)
- [Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting](https://mlsanity.com/paper/2310.11324/) (2023)
- [From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge](https://mlsanity.com/paper/2411.16594/) (2024)
- [Humanity's Last Exam](https://mlsanity.com/paper/2501.14249/) (2025)
- [LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods](https://mlsanity.com/paper/2412.05579/) (2024)
- [Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models](https://mlsanity.com/paper/2404.18796/) (2024)
- [Omni-MATH: A Universal Olympiad Level Mathematic Benchmark For Large Language Models](https://mlsanity.com/paper/2410.07985/) (2024)
- [FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI](https://mlsanity.com/paper/2411.04872/) (2024)
- [Are We Done with MMLU?](https://mlsanity.com/paper/2406.04127/) (2024)
- [Lessons from the Trenches on Reproducible Evaluation of Language Models](https://mlsanity.com/paper/2405.14782/) (2024)
- [Adding Error Bars to Evals: A Statistical Approach to Language Model Evaluations](https://mlsanity.com/paper/2411.00640/) (2024)
- [BetterBench: Assessing AI Benchmarks, Uncovering Issues, and Establishing Best Practices](https://mlsanity.com/paper/2411.12990/) (2024)
- [Open-LLM-Leaderboard: From Multi-choice to Open-style Questions for LLMs Evaluation, Benchmark, and Arena](https://mlsanity.com/paper/2406.07545/) (2024)
- [The Alternative Annotator Test for LLM-as-a-Judge: How to Statistically Justify Replacing Human Annotators with LLMs](https://mlsanity.com/paper/2501.10970/) (2025)
- [Can You Trust LLM Judgments? Reliability of LLM-as-a-Judge](https://mlsanity.com/paper/2412.12509/) (2024)
- [Do Large Language Model Benchmarks Test Reliability?](https://mlsanity.com/paper/2502.03461/) (2025)
- [Large Language Model Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation](https://mlsanity.com/paper/2509.08825/) (2025)
- [Auto-Arena: Automating LLM Evaluations with Agent Peer Battles and Committee Discussions](https://mlsanity.com/paper/2405.20267/) (2024)
- [Position: Don't Use the CLT in LLM Evals With Fewer Than a Few Hundred Datapoints](https://mlsanity.com/paper/2503.01747/) (2025)
- [On Robustness and Reliability of Benchmark-Based Evaluation of LLMs](https://mlsanity.com/paper/2509.04013/) (2025)
- [xFinder: Large Language Models as Automated Evaluators for Reliable Evaluation](https://mlsanity.com/paper/2405.11874/) (2024)
- [JudgeBlender: Ensembling Judgments for Automatic Relevance Assessment](https://mlsanity.com/paper/2412.13268/) (2024)

## Built on by (most cited first, 11 of 11)

- [BenchGuard: Who Guards the Benchmarks? Automated Auditing of LLM Agent Benchmarks](https://mlsanity.com/paper/2604.24955/) (2026)
- [Precise but Uncoupled: Reviewer Precision Does Not Guarantee Critique Uptake in Multi-Agent Math Reasoning](https://mlsanity.com/paper/2607.15388/) (2026)
- [How Much Does a Reasoning Summary Reveal? An Observability Ladder for Large Language Models](https://mlsanity.com/paper/2608.02089/) (2026)
- [Shorter Reasoning, Earlier Answers? An Evaluation of Reasoning Interfaces](https://mlsanity.com/paper/2608.03401/) (2026)
- [Wrong but Useful: Trajectory Value Beyond Answer Correctness in Multi-Agent Messages](https://mlsanity.com/paper/2608.14375/) (2026)
- [Early Evidence of Vibe-Proving with Consumer LLMs: A Case Study on Spectral Region Characterization with ChatGPT-5.2 (Thinking)](https://mlsanity.com/paper/2602.18918/) (2026)
- [Thinking Like a Scientist? A Structural Study of LLM-Generated Research Methods](https://mlsanity.com/paper/2606.26130/) (2026)
- [LLMs Can Predict Failure Risk, But Struggle to Predict Which Collaboration Protocol Pays Off: Cost-Aware Protocol Routing Across Reasoning Tasks](https://mlsanity.com/paper/2608.14927/) (2026)
- [Beyond Repeated Sampling: Learning Search Policies for LLM Reasoning](https://mlsanity.com/paper/2609.26704/) (2026)
- [Universal Textual Teaching for LLMs](https://mlsanity.com/paper/2610.12114/) (2026)
- [Big-Big-SycoBench: A Benchmark to Detect Sycophancy Through Model Responses to Valid vs. Invalid Arguments](https://mlsanity.com/paper/2610.15634/) (2026)

## Abstract (arXiv)

Benchmarks are important tools to track progress in the development of Large Language Models (LLMs), yet inaccuracies in datasets and evaluation methods consistently undermine their effectiveness. Here, we present Omni-MATH-2, a manually revised version of the Omni-MATH dataset comprising a clean, exact-answer subset (n=4181) and a tagged, non-standard subset (n=247). Each problem was audited to ensure LaTeX compilability, solvability and verifiability, which involved adding missing figures or information, labeling problems requiring a proof, estimation or image, and removing clutter. This process significantly reduces dataset-induced noise, thereby providing a more precise assessment of model performance. The annotated dataset also allows us to evaluate judge-induced noise by comparing GPT-5 mini with the original Omni-Judge, revealing substantial discrepancies between judges on both the clean and tagged problem subsets. Expert annotations reveal that Omni-Judge is wrong in 96.4% of the judge disagreements, indicating its inability to differentiate between models' abilities, even well before saturation of the benchmark occurs. As problems become more challenging, we find that increasingly competent judges become essential in order to prevent judge errors from masking genuine differences between models. Finally, neither judge identifies the present failure modes for the subset of tagged problems, demonstrating that dataset quality and judge reliability are both critical to develop accurate benchmarks of model performance.