# When LLMs Benchmark Themselves: Deconstructing Self-Bias in Automated Evaluation

arXiv: https://arxiv.org/abs/2509.26600 · HTML: https://mlsanity.com/paper/2509.26600/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 15 of its 26 references that are in the map.
- First posted on arXiv 30 Sep 2025.

## Builds on (most cited first, 15 of 15)

- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [LLM Evaluators Recognize and Favor Their Own Generations](https://mlsanity.com/paper/2404.13076/) (2024)
- [From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline](https://mlsanity.com/paper/2406.11939/) (2024)
- [Large Language Models Are State-of-the-Art Evaluators of Translation Quality](https://mlsanity.com/paper/2302.14520/) (2023)
- [Self-Preference Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2410.21819/) (2024)
- [Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Bias](https://mlsanity.com/paper/2306.15895/) (2023)
- [MuSR: Testing the Limits of Chain-of-thought with Multistep Soft Reasoning](https://mlsanity.com/paper/2310.16049/) (2023)
- [Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement](https://mlsanity.com/paper/2402.11436/) (2024)
- [No Free Labels: Limitations of LLM-as-a-Judge Without Human Grounding](https://mlsanity.com/paper/2503.05061/) (2025)
- [Do LLM Evaluators Prefer Themselves for a Reason?](https://mlsanity.com/paper/2504.03846/) (2025)
- [Efficacy of Synthetic Data as a Benchmark](https://mlsanity.com/paper/2409.11968/) (2024)
- [Automated Benchmark Generation for Repository-Level Coding Tasks](https://mlsanity.com/paper/2503.07701/) (2025)
- [Do These LLM Benchmarks Agree? Fixing Benchmark Evaluation with BenchBench](https://mlsanity.com/paper/2407.13696/) (2024)
- [Zero-shot Benchmarking: A Framework for Flexible and Scalable Automatic Evaluation of Language Models](https://mlsanity.com/paper/2504.01001/) (2025)
- [Automatic Generation of Benchmarks and Reliable LLM Judgment for Code Tasks](https://mlsanity.com/paper/2410.21071/) (2024)

## Built on by (most cited first, 3 of 3)

- [GaelEval: Benchmarking LLM Performance for Scottish Gaelic](https://mlsanity.com/paper/2604.02135/) (2026)
- [Evaluating Commercial AI Chatbots as News Intermediaries](https://mlsanity.com/paper/2605.22785/) (2026)
- [Last Translation Benchmark](https://mlsanity.com/paper/2609.04173/) (2026)

## Abstract (arXiv)

As LLMs rapidly saturate existing benchmarks, automated benchmark creation using LLMs (LLM as a benchmark) where a model generates test inputs (LLM as a testset) and evaluates outputs (LLM as an evaluator) has gained traction as a cheap alternative to human curation. We show that this paradigm has a fundamental problem: LLM-generated benchmarks systematically favor the model that created them. Using machine translation as our primary testbed, we find that self bias arises from two compounding sources, LLM as a testset and LLM as an evaluator, and their combination amplifies the effect. Crucially, even when test data is generated with explicit diversity controls, each model's implicit stylistic tendencies produce homogeneous, model-specific outputs that inflate its own scores. Increasing source text diversity, using our proposed diversity metric, partially mitigates this bias. Self bias is strong enough to cause each model to rank itself first, overriding the peer consensus ordering. We confirm that the phenomenon extends to open-ended generation on the Chatbot Arena task.