# CARE: Confounder-Aware Aggregation for Reliable LLM Evaluation

arXiv: https://arxiv.org/abs/2603.00039 · HTML: https://mlsanity.com/paper/2603.00039/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Builds on 49 of its 72 references that are in the map.
- First posted on arXiv 9 Feb 2026.

## Builds on (most cited first, 30 of 49)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [Large Language Models are not Fair Evaluators](https://mlsanity.com/paper/2305.17926/) (2023)
- [Can Large Language Models Be an Alternative to Human Evaluations?](https://mlsanity.com/paper/2305.01937/) (2023)
- [Yi: Open Foundation Models by 01.AI](https://mlsanity.com/paper/2403.04652/) (2024)
- [Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs](https://mlsanity.com/paper/2503.01743/) (2025)
- [RewardBench: Evaluating Reward Models for Language Modeling](https://mlsanity.com/paper/2403.13787/) (2024)
- [LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods](https://mlsanity.com/paper/2412.05579/) (2024)
- [Large Language Models Are State-of-the-Art Evaluators of Translation Quality](https://mlsanity.com/paper/2302.14520/) (2023)
- [Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification](https://mlsanity.com/paper/1903.04561/) (2019)
- [Snorkel: Rapid Training Data Creation with Weak Supervision](https://mlsanity.com/paper/1711.10160/) (2017)
- [Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge](https://mlsanity.com/paper/2410.02736/) (2024)
- [JudgeLM: Fine-tuned Large Language Models are Scalable Judges](https://mlsanity.com/paper/2310.17631/) (2023)
- [Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models](https://mlsanity.com/paper/2404.18796/) (2024)
- [Humans or LLMs as the Judge? A Study on Judgement Biases](https://mlsanity.com/paper/2402.10669/) (2024)
- [PandaLM: An Automatic Evaluation Benchmark for LLM Instruction Tuning Optimization](https://mlsanity.com/paper/2306.05087/) (2023)
- [Evaluating Large Language Models at Evaluating Instruction Following](https://mlsanity.com/paper/2310.07641/) (2023)
- [Understanding Dataset Difficulty with V-Usable Information](https://mlsanity.com/paper/2110.08420/) (2021)
- [Debating with More Persuasive LLMs Leads to More Truthful Answers](https://mlsanity.com/paper/2402.06782/) (2024)
- [PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference](https://mlsanity.com/paper/2406.15513/) (2024)
- [Crossing the Reward Bridge: Expanding RL with Verifiable Rewards Across Diverse Domains](https://mlsanity.com/paper/2503.23829/) (2025)
- [CycleResearcher: Improving Automated Research via Automated Review](https://mlsanity.com/paper/2411.00816/) (2024)
- [Preference Leakage: A Contamination Problem in LLM-as-a-judge](https://mlsanity.com/paper/2502.01534/) (2025)

## Built on by (most cited first, 13 of 13)

- [Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines](https://mlsanity.com/paper/2604.23178/) (2026)
- [ParetoBandit: Budget-Paced Adaptive Routing for Non-Stationary LLM Serving](https://mlsanity.com/paper/2604.00136/) (2026)
- [CoEval: Ranking Language Models for Custom Tasks Without Labeled Data or Trustworthy Benchmarks](https://mlsanity.com/paper/2606.03650/) (2026)
- [Are Verifier Errors Independent Within a GRPO Group? Evidence from Qwen2.5 Rollouts](https://mlsanity.com/paper/2609.06386/) (2026)
- [LLM-FACETS: A Privacy-Preserving Framework for Evaluating LLM Transparency and Accountability](https://mlsanity.com/paper/2605.31167/) (2026)
- [Codifying the Judge: Scalable Evaluation via Program Distillation](https://mlsanity.com/paper/2607.22561/) (2026)
- [Messier: A High-Resolution Corpus for Cross-Benchmark Agent Evaluation](https://mlsanity.com/paper/2607.25891/) (2026)
- [Conditional Evaluation of Language Models with Cheap Auxiliary Signals](https://mlsanity.com/paper/2608.16210/) (2026)
- [A Closed-Form Estimator and Diagnostic Battery for Anchor-Judge Error Correlation, Under a Single-Common-Factor Model](https://mlsanity.com/paper/2609.08826/) (2026)
- [Observational Equivalence of LLM and Human Annotation](https://mlsanity.com/paper/2609.22133/) (2026)
- [RAIM: Robust Aggregation of Inexpensive Models for Hallucination Detection](https://mlsanity.com/paper/2609.39229/) (2026)
- [Can We Trust This Evaluation? Instance-Level Reliability of LLM-as-a-Judge](https://mlsanity.com/paper/2610.14029/) (2026)
- [Checking the Checkers: A Synthesis of Verifier Noise and Robustness in RLVR](https://mlsanity.com/paper/2610.17824/) (2026)

## Abstract (arXiv)

LLM-as-a-judge ensembles are the standard paradigm for scalable evaluation, but their aggregation mechanisms suffer from a fundamental flaw: they implicitly assume that judges provide independent estimates of true quality. However, in practice, LLM judges exhibit correlated errors caused by shared latent confounders -- such as verbosity, stylistic preferences, or training artifacts -- causing standard aggregation rules like majority vote or averaging to provide little gain or even amplify systematic mistakes. To address this, we introduce CARE, a confounder-aware aggregation framework that explicitly models LLM judge scores as arising from both a latent true-quality signal and shared confounding factors. Rather than heuristically re-weighting judges, CARE separates quality from confounders without access to ground-truth labels. We provide theoretical guarantees for identifiability and finite-sample recovery under shared confounders, and we quantify the systematic bias incurred when aggregation models omit confounding latent factors. Across 12 public benchmarks spanning continuous scoring, binary classification, and pairwise preference settings, CARE improves aggregation accuracy, reducing error by up to 26.8%. Code is released in hrefhttps://github.com/SprocketLab/CAREhttps://github.com/SprocketLab/CARE.