# Boundaries Agree, Labels Do Not: Intra-Annotator Dynamics as a Kind of Training Data

arXiv: https://arxiv.org/abs/2610.04370 · HTML: https://mlsanity.com/paper/2610.04370/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 28 references that are in the map.
- First posted on arXiv 3 Oct 2026.

## Builds on (most cited first, 7 of 7)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Artificial Artificial Artificial Intelligence: Crowd Workers Widely Use Large Language Models for Text Production Tasks](https://mlsanity.com/paper/2306.07899/) (2023)
- [Reliability without Validity: A Systematic, Large-Scale Evaluation of LLM-as-a-Judge Models Across Agreement, Consistency, and Bias](https://mlsanity.com/paper/2606.19544/) (2026)
- [Consistency is Key: Disentangling Label Variation in Natural Language Processing with Intra-Annotator Agreement](https://mlsanity.com/paper/2301.10684/) (2023)
- [Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLP](https://mlsanity.com/paper/2601.09065/) (2026)
- [NUTMEG: Separating Signal From Noise in Annotator Disagreement](https://mlsanity.com/paper/2507.18890/) (2025)
- [Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization](https://mlsanity.com/paper/2605.28802/) (2026)

## Abstract (arXiv)

Data quality now matters as much as compute for training language models. Much training data comes from human annotation of text, and interpretive annotation has no ground truth that could settle what is "accurate". Two lines of work respond to this. One combines annotators into a "ground truth" and measures how well they agree with each other; the other treats their disagreement as a signal. Both compare different people at one point in time. We measure something else: how well one reader reproduces their own reading of the same text over time. One expert human reader and three LLM families segmented three Sumerian myths and labelled the causal function of each segment with one of seven states. Across runs months apart, the human cut the text in much the same places but named the segments differently, in every myth. The models show no such consistent pattern: their gap between the two layers is positive in some myths and negative in others, and its size varies. The human's label changes are not random: the runs go through much the same functions but start them one step apart, while model runs start them at the same places. We argue that this pattern is a usable measure of data quality and a contamination check: a "human" annotation whose labels are as stable as its boundaries, and whose functions start in sync, looks like a model's.