# Spying on your neighbors: Fine-grained probing of contextual embeddings for information about surrounding words

arXiv: https://arxiv.org/abs/2005.01810 · HTML: https://mlsanity.com/paper/2005.01810/

## Key facts

- Cited by 28 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 4 May 2020.

## Built on by (most cited first, 28 of 28)

- [Probing for the Usage of Grammatical Number](https://mlsanity.com/paper/2204.08831/) (2022)
- [Local Interpretations for Explainable Natural Language Processing: A Survey](https://mlsanity.com/paper/2103.11072/) (2021)
- [Exploring the Role of BERT Token Representations to Explain Sentence Probing Results](https://mlsanity.com/paper/2104.01477/) (2021)
- [Sorting through the noise: Testing robustness of information processing in pre-trained language models](https://mlsanity.com/paper/2109.12393/) (2021)
- [Holmes: A Benchmark to Assess the Linguistic Competence of Language Models](https://mlsanity.com/paper/2404.18923/) (2024)
- [Neural reality of argument structure constructions](https://mlsanity.com/paper/2202.12246/) (2022)
- [AdapLeR: Speeding up Inference by Adaptive Length Reduction](https://mlsanity.com/paper/2203.08991/) (2022)
- [On the Importance of Data Size in Probing Fine-tuned Models](https://mlsanity.com/paper/2203.09627/) (2022)
- [Chunk, Align, Select: A Simple Long-sequence Processing Method for Transformers](https://mlsanity.com/paper/2308.13191/) (2023)
- [Towards a Similarity-adjusted Surprisal Theory](https://mlsanity.com/paper/2410.17676/) (2024)
- [Assessing the Capacity of Transformer to Abstract Syntactic Representations: A Contrastive Analysis Based on Long-distance Agreement](https://mlsanity.com/paper/2212.04523/) (2022)
- [Linguistic Interpretability of Transformer-based Language Models: a systematic review](https://mlsanity.com/paper/2504.08001/) (2025)
- [On the Interplay Between Fine-tuning and Composition in Transformers](https://mlsanity.com/paper/2105.14668/) (2021)
- [Learning Sense-Specific Static Embeddings using Contextualised Word Embeddings as a Proxy](https://mlsanity.com/paper/2110.02204/) (2021)
- [Representing Affect Information in Word Embeddings](https://mlsanity.com/paper/2209.10583/) (2022)
- [Automatic Scoring of Dream Reports' Emotional Content with Large Language Models](https://mlsanity.com/paper/2302.14828/) (2023)
- [From Transcripts to Insights: Uncovering Corporate Risks Using Generative AI](https://mlsanity.com/paper/2310.17721/) (2023)
- [Understanding Subword Compositionality of Large Language Models](https://mlsanity.com/paper/2508.17953/) (2025)
- [Variation and generality in encoding of syntactic anomaly information in sentence embeddings](https://mlsanity.com/paper/2111.06644/) (2021)
- [Reconstruction Probing](https://mlsanity.com/paper/2212.10792/) (2022)
- [Steering Language Models in Multi-Token Generation: A Case Study on Tense and Aspect](https://mlsanity.com/paper/2509.12065/) (2025)
- [Counteracts: Testing Stereotypical Representation in Pre-trained Language Models](https://mlsanity.com/paper/2301.04347/) (2023)
- [Do Pretrained Contextual Language Models Distinguish between Hebrew Homograph Analyses?](https://mlsanity.com/paper/2405.07099/) (2024)
- [A Representation Level Analysis of NMT Model Robustness to Grammatical Errors](https://mlsanity.com/paper/2505.21224/) (2025)
- [Learning to Look at the Other Side: A Semantic Probing Study of Word Embeddings in LLMs with Enabled Bidirectional Attention](https://mlsanity.com/paper/2510.01652/) (2025)
- [The Grammar of Transformers: A Systematic Review of Interpretability Research on Syntactic Knowledge in Language Models](https://mlsanity.com/paper/2601.19926/) (2026)
- [Probing structural constraints of negation in Pretrained Language Models](https://mlsanity.com/paper/2408.03070/) (2024)
- [Exposure is Optional: Learning Unlike Coordination in Language Models](https://mlsanity.com/paper/2607.20251/) (2026)

## Abstract (arXiv)

Although models using contextual word embeddings have achieved state-of-the-art results on a host of NLP tasks, little is known about exactly what information these embeddings encode about the context words that they are understood to reflect. To address this question, we introduce a suite of probing tasks that enable fine-grained testing of contextual embeddings for encoding of information about surrounding words. We apply these tasks to examine the popular BERT, ELMo and GPT contextual encoders, and find that each of our tested information types is indeed encoded as contextual information across tokens, often with near-perfect recoverability-but the encoders vary in which features they distribute to which tokens, how nuanced their distributions are, and how robust the encoding of each feature is to distance. We discuss implications of these results for how different types of models breakdown and prioritize word-level context information when constructing token embeddings.