# A Controlled Reevaluation of Coreference Resolution Models

arXiv: https://arxiv.org/abs/2404.00727 · HTML: https://mlsanity.com/paper/2404.00727/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 42 references that are in the map.
- First posted on arXiv 31 Mar 2024.

## Builds on (most cited first, 22 of 22)

- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://mlsanity.com/paper/2006.03654/) (2020)
- [mT5: A massively multilingual pre-trained text-to-text transformer](https://mlsanity.com/paper/2010.11934/) (2020)
- [SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems](https://mlsanity.com/paper/1905.00537/) (2019)
- [End-to-end Neural Coreference Resolution](https://mlsanity.com/paper/1707.07045/) (2017)
- [Higher-order Coreference Resolution with Coarse-to-fine Inference](https://mlsanity.com/paper/1804.05392/) (2018)
- [BERT for Coreference Resolution: Baselines and Analysis](https://mlsanity.com/paper/1908.09091/) (2019)
- [On the State of the Art of Evaluation in Neural Language Models](https://mlsanity.com/paper/1707.05589/) (2017)
- [Revealing the Myth of Higher-Order Inference in Coreference Resolution](https://mlsanity.com/paper/2009.12013/) (2020)
- [Coreference Resolution without Span Representations](https://mlsanity.com/paper/2101.00434/) (2021)
- [Autoregressive Structured Prediction with Language Models](https://mlsanity.com/paper/2210.14698/) (2022)
- [Word-Level Coreference Resolution](https://mlsanity.com/paper/2109.04127/) (2021)
- [Coreference Resolution through a seq2seq Transition-Based System](https://mlsanity.com/paper/2211.12142/) (2022)
- [LingMess: Linguistically Informed Multi Expert Scorers for Coreference Resolution](https://mlsanity.com/paper/2205.12644/) (2022)
- [On Generalization in Coreference Resolution](https://mlsanity.com/paper/2109.09667/) (2021)
- [A Cross-Task Analysis of Text Span Representations](https://mlsanity.com/paper/2006.03866/) (2020)
- [OntoGUM: Evaluating Contextualized SOTA Coreference Resolution on 12 More Genres](https://mlsanity.com/paper/2106.00933/) (2021)
- [Findings of the Shared Task on Multilingual Coreference Resolution](https://mlsanity.com/paper/2209.07841/) (2022)
- [End-to-end Neural Coreference Resolution Revisited: A Simple yet Effective Baseline](https://mlsanity.com/paper/2107.01700/) (2021)
- [On "Scientific Debt" in NLP: A Case for More Rigour in Language Model Pre-Training Research](https://mlsanity.com/paper/2306.02870/) (2023)
- [Challenges to Evaluating the Generalization of Coreference Resolution Models: A Measurement Modeling Perspective](https://mlsanity.com/paper/2303.09092/) (2023)
- [Anatomy of OntoGUM--Adapting GUM to the OntoNotes Scheme to Evaluate Robustness of SOTA Coreference Algorithms](https://mlsanity.com/paper/2110.05727/) (2021)

## Built on by (most cited first, 4 of 4)

- [The Elephant in the Coreference Room: Resolving Coreference in Full-Length French Fiction Works](https://mlsanity.com/paper/2510.15594/) (2025)
- [Just Use XML: Revisiting Joint Translation and Label Projection](https://mlsanity.com/paper/2603.12021/) (2026)
- [Closing the Gap at CRAC 2026: Two-Stage Adaptation for LLM-Based Multilingual Coreference Resolution](https://mlsanity.com/paper/2605.16984/) (2026)
- [On the use of foundation models in cognitive science](https://mlsanity.com/paper/2608.07812/) (2026)

## Abstract (arXiv)

All state-of-the-art coreference resolution (CR) models involve finetuning a pretrained language model. Whether the superior performance of one CR model over another is due to the choice of language model or other factors, such as the task-specific architecture, is difficult or impossible to determine due to lack of a standardized experimental setup. To resolve this ambiguity, we systematically evaluate five CR models and control for certain design decisions including the pretrained language model used by each. When controlling for language model size, encoder-based CR models outperform more recent decoder-based models in terms of both accuracy and inference speed. Surprisingly, among encoder-based CR models, more recent models are not always more accurate, and the oldest CR model that we test generalizes the best to out-of-domain textual genres. We conclude that controlling for the choice of language model reduces most, but not all, of the increase in F1 score reported in the past five years.