# In the LLM era, Word Sense Induction remains unsolved

arXiv: https://arxiv.org/abs/2603.11686 · HTML: https://mlsanity.com/paper/2603.11686/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 15 of its 50 references that are in the map.
- First posted on arXiv 12 Mar 2026.

## Builds on (most cited first, 15 of 15)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [WiC: the Word-in-Context Dataset for Evaluating Context-Sensitive Meaning Representations](https://mlsanity.com/paper/1808.09121/) (2018)
- [SemEval-2020 Task 1: Unsupervised Lexical Semantic Change Detection](https://mlsanity.com/paper/2007.11464/) (2020)
- [Towards better substitution-based word sense induction](https://mlsanity.com/paper/1905.12598/) (2019)
- [MirrorWiC: On Eliciting Word-in-Context Representations from Pretrained Language Models](https://mlsanity.com/paper/2109.09237/) (2021)
- [FEWS: Large-Scale, Low-Shot Word Sense Disambiguation with the Dictionary](https://mlsanity.com/paper/2102.07983/) (2021)
- [Semantic Frame Induction using Masked Word Embeddings and Two-Step Clustering](https://mlsanity.com/paper/2105.13466/) (2021)
- [Large Scale Substitution-based Word Sense Induction](https://mlsanity.com/paper/2110.07681/) (2021)
- [What do Language Models know about word senses? Zero-Shot WSD with Language Models and Domain Inventories](https://mlsanity.com/paper/2302.03353/) (2023)
- [AutoSense Model for Word Sense Induction](https://mlsanity.com/paper/1811.09242/) (2018)
- [Unsupervised Word Polysemy Quantification with Multiresolution Grids of Contextual Embeddings](https://mlsanity.com/paper/2003.10224/) (2020)
- [PolyLM: Learning about Polysemy through Language Modeling](https://mlsanity.com/paper/2101.10448/) (2021)
- [To Word Senses and Beyond: Inducing Concepts with Contextualized Language Models](https://mlsanity.com/paper/2406.20054/) (2024)
- [Injecting Wiktionary to improve token-level contextual representations using contrastive learning](https://mlsanity.com/paper/2402.07817/) (2024)

## Built on by (most cited first, 2 of 2)

- [Relation Geometry in Semantic Space of Language Models](https://mlsanity.com/paper/2607.26762/) (2026)
- [Technical Manual for a Toolkit for Measuring Contextual Individuation in Transformer Language Models](https://mlsanity.com/paper/2609.05333/) (2026)

## Abstract (arXiv)

In the absence of sense-annotated data, word sense induction (WSI) is a compelling alternative to word sense disambiguation, particularly in low-resource or domain-specific settings. In this paper, we emphasize methodological problems in current WSI evaluation. We propose an evaluation on a SemCor-derived dataset, respecting the original corpus polysemy and frequency distributions. We assess pre-trained embeddings and clustering algorithms across parts of speech, and propose and evaluate an LLM-based WSI method for English. We evaluate data augmentation sources (LLM-generated, corpus and lexicon), and semi-supervised scenarios using Wiktionary for data augmentation, must-link constraints, number of clusters per lemma. We find that no unsupervised method (whether ours or previous) surpasses the strong "one cluster per lemma" heuristic (1cpl). We also show that (i) results and best systems may vary across POS, (ii) LLMs have troubles performing this task, (iii) data augmentation is beneficial and (iv) capitalizing on Wiktionary does help. It surpasses previous SOTA system on our test set by 3.3%. WSI is not solved, and calls for a better articulation of lexicons and LLMs' lexical semantics capabilities.