Towards Provably Unbiased LLM Judges via Bias-Bounded Evaluation
Not on the map. Why?
Abstract (arXiv)
As AI models progress beyond simple chatbots into more complex workflows, we draw ever closer to the event horizon beyond which AI systems will be utilized in autonomous, self-maintaining feedback loops. Any autonomous AI system will depend on automated, verifiable rewards and feedback; in settings where ground truth is sparse or non-deterministic, one practical source of such rewards is an LLM-as-a-Judge. Although LLM judges continue to improve, the literature has yet to introduce systems capable of enforcing standards with strong guarantees, particularly when bias vectors are unknown or adversarially discovered. To remedy this issue, we propose average bias-boundedness (A-BB), an algorithmic framework which formally guarantees reductions of harm/impact as a result of any measurable bias in an LLM judge. Evaluating on Arena-Hard-Auto with four LLM judges, we achieve (tau=0.5, delta=0.01) bias-bounded guarantees while retaining 61-99% correlation with original rankings across formatting and schematic bias settings, with most judge-bias combinations exceeding 80%. The code to reproduce our findings is available at https://github.com/penfever/bias-bounded-evaluation.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 20 of its 39 references that are in the map. First posted on arXiv 5 Mar 2026.
Builds on
20 of 39 references.
References · 20
- 2024Qwen2.5 Technical Reportcited by 4,713
- 2021Convex optimizationcited by 2,479
- 2019Introduction to Online Convex Optimizationcited by 793
- 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 778
- 2024From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipelinecited by 484
- 2024Justice or Prejudice? Quantifying Biases in LLM-as-a-Judgecited by 354
- 2024JudgeBench: A Benchmark for Evaluating LLM-based Judgescited by 294
- 2013The Composition Theorem for Differential Privacycited by 290
- 2024Agent-as-a-Judge: Evaluate Agents with Agentscited by 223
- 2018Improving the Gaussian Mechanism for Differential Privacy: Analytical Calibration and Optimal Denoisingcited by 216
- 2025Preference Leakage: A Contamination Problem in LLM-as-a-judgecited by 119
- 2025Agentic Large Language Models, a surveycited by 105
8 more
- 2024Trust or Escalate: LLM Judges with Provable Guarantees for Human Agreementcited by 99
- 2025Evaluating Scoring Bias in LLM-as-a-Judgecited by 42
- 2024Style Outweighs Substance: Failure Modes of LLM Judges in Alignment Benchmarkingcited by 35
- 2023Advancing Differential Privacy: Where We Are Now and Future Directions for Real-World Deploymentcited by 26
- 2025Analyzing Uncertainty of LLM-as-a-Judge: Interval Evaluations with Conformal Predictioncited by 26
- 2025Beyond Consensus: Mitigating the Agreeableness Bias in LLM Judge Evaluationscited by 16
- 2025When Judgment Becomes Noise: How Design Failures in LLM Judge Benchmarks Silently Undermine Validitycited by 11
- 2025Calibrating LLM Judges: Linear Probes for Fast and Reliable Uncertainty Estimationcited by 6
Built on by
Citing papers per year: 2026: 1.
1 paper.