# NUTMEG: Separating Signal From Noise in Annotator Disagreement

arXiv: https://arxiv.org/abs/2507.18890 · HTML: https://mlsanity.com/paper/2507.18890/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 52 references that are in the map.
- First posted on arXiv 25 Jul 2025.

## Builds on (most cited first, 21 of 21)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Optuna: A Next-generation Hyperparameter Optimization Framework](https://mlsanity.com/paper/1907.10902/) (2019)
- [Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference](https://mlsanity.com/paper/2412.13663/) (2024)
- [Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotations](https://mlsanity.com/paper/2110.05719/) (2021)
- [Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detection](https://mlsanity.com/paper/2111.07997/) (2021)
- [Toward a Perspectivist Turn in Ground Truthing for Predictive Computing](https://mlsanity.com/paper/2109.04270/) (2021)
- [NLPositionality: Characterizing Design Biases of Datasets and Models](https://mlsanity.com/paper/2306.01943/) (2023)
- [When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks](https://mlsanity.com/paper/2305.06626/) (2023)
- [Designing Toxic Content Classification for a Diversity of Perspectives](https://mlsanity.com/paper/2106.04511/) (2021)
- [Everyone's Voice Matters: Quantifying Annotation Disagreement Using Demographic Information](https://mlsanity.com/paper/2301.05036/) (2023)
- [When Do Annotator Demographics Matter? Measuring the Influence of Annotator Demographics with the POPQUORN Dataset](https://mlsanity.com/paper/2306.06826/) (2023)
- [The Ecological Fallacy in Annotation: Modelling Human Label Variation goes beyond Sociodemographics](https://mlsanity.com/paper/2306.11559/) (2023)
- [VariErr NLI: Separating Annotation Error from Human Label Variation](https://mlsanity.com/paper/2403.01931/) (2024)
- [You Are What You Annotate: Towards Better Models through Annotator Representations](https://mlsanity.com/paper/2305.14663/) (2023)
- [The Perspectivist Paradigm Shift: Assumptions and Challenges of Capturing Human Labels](https://mlsanity.com/paper/2405.05860/) (2024)
- [Capturing Perspectives of Crowdsourced Annotators in Subjective Learning Tasks](https://mlsanity.com/paper/2311.09743/) (2023)
- [CrowdTruth 2.0: Quality Metrics for Crowdsourcing with Disagreement](https://mlsanity.com/paper/1808.06080/) (2018)
- [ACTOR: Active Learning with Annotator-specific Classification Heads to Embrace Human Label Variation](https://mlsanity.com/paper/2310.14979/) (2023)
- [CROWDLAB: Supervised learning to infer consensus labels and quality scores for data with multiple annotators](https://mlsanity.com/paper/2210.06812/) (2022)
- [Voices in a Crowd: Searching for Clusters of Unique Perspectives](https://mlsanity.com/paper/2407.14259/) (2024)
- [Whose Preferences? Differences in Fairness Preferences and Their Impact on the Fairness of AI Utilizing Human Feedback](https://mlsanity.com/paper/2406.05902/) (2024)

## Built on by (most cited first, 7 of 7)

- [Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLP](https://mlsanity.com/paper/2601.09065/) (2026)
- [Modeling Annotator Disagreement with Demographic-Aware Experts and Synthetic Perspectives](https://mlsanity.com/paper/2508.02853/) (2025)
- [Are we chasing ghosts? Quantifying unattributable polarization, and attributing the rest to annotator groups](https://mlsanity.com/paper/2602.06055/) (2026)
- [Not All Subjectivity Is the Same! Defining Desiderata for the Evaluation of Subjectivity in NLP](https://mlsanity.com/paper/2603.28351/) (2026)
- [LLM-Assisted Stance Detection in Scientific Discourse: A Test Case in Bayesian Cognitive Science](https://mlsanity.com/paper/2606.15566/) (2026)
- [Boundaries Agree, Labels Do Not: Intra-Annotator Dynamics as a Kind of Training Data](https://mlsanity.com/paper/2610.04370/) (2026)
- [Whose Ground Truth? Embracing Ambiguity in Human-Centered AI](https://mlsanity.com/paper/2610.10805/) (2026)

## Abstract (arXiv)

NLP models often rely on human-labeled data for training and evaluation. Many approaches crowdsource this data from a large number of annotators with varying skills, backgrounds, and motivations, resulting in conflicting annotations. These conflicts have traditionally been resolved by aggregation methods that assume disagreements are errors. Recent work has argued that for many tasks annotators may have genuine disagreements and that variation should be treated as signal rather than noise. However, few models separate signal and noise in annotator disagreement. In this work, we introduce NUTMEG, a new Bayesian model that incorporates information about annotator backgrounds to remove noisy annotations from human-labeled training data while preserving systematic disagreements. Using synthetic data, we show that NUTMEG is more effective at recovering ground-truth from annotations with systematic disagreement than traditional aggregation methods. We provide further analysis characterizing how differences in subpopulation sizes, rates of disagreement, and rates of spam affect the performance of our model. Finally, we demonstrate that downstream models trained on NUTMEG-aggregated data significantly outperform models trained on data from traditionally aggregation methods. Our results highlight the importance of accounting for both annotator competence and systematic disagreements when training on human-labeled data.