# Language model acceptability judgements are not always robust to context

arXiv: https://arxiv.org/abs/2212.08979 · HTML: https://mlsanity.com/paper/2212.08979/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- Builds on 41 of its 61 references that are in the map.
- First posted on arXiv 18 Dec 2022.

## Builds on (most cited first, 30 of 41)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Making Pre-trained Language Models Better Few-shot Learners](https://mlsanity.com/paper/2012.15723/) (2020)
- [Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity](https://mlsanity.com/paper/2104.08786/) (2021)
- [Neural Network Acceptability Judgments](https://mlsanity.com/paper/1805.12471/) (2018)
- [Exploiting Cloze Questions for Few Shot Text Classification and Natural Language Inference](https://mlsanity.com/paper/2001.07676/) (2020)
- [CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models](https://mlsanity.com/paper/2010.00133/) (2020)
- [It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners](https://mlsanity.com/paper/2009.07118/) (2020)
- [BLiMP: The Benchmark of Linguistic Minimal Pairs for English](https://mlsanity.com/paper/1912.00582/) (2019)
- [Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies](https://mlsanity.com/paper/1611.01368/) (2016)
- [Do Prompt-Based Models Really Understand the Meaning of their Prompts?](https://mlsanity.com/paper/2109.01247/) (2021)
- [Compositionality decomposed: how do neural networks generalise?](https://mlsanity.com/paper/1908.08351/) (2019)
- [Can language models learn from explanations in context?](https://mlsanity.com/paper/2204.02329/) (2022)
- [Negated and Misprimed Probes for Pretrained Language Models: Birds Can Talk, But Cannot Fly](https://mlsanity.com/paper/1911.03343/) (2019)
- [Targeted Syntactic Evaluation of Language Models](https://mlsanity.com/paper/1808.09031/) (2018)
- [Language models show human-like content effects on reasoning tasks](https://mlsanity.com/paper/2207.07051/) (2022)
- [A Systematic Assessment of Syntactic Generalization in Neural Language Models](https://mlsanity.com/paper/2005.03692/) (2020)
- [Transformer Language Models without Positional Encodings Still Learn Positional Information](https://mlsanity.com/paper/2203.16634/) (2022)
- [State-of-the-art generalisation research in NLP: A taxonomy and review](https://mlsanity.com/paper/2210.03050/) (2022)
- [Are Natural Language Inference Models IMPPRESsive? Learning IMPlicature and PRESupposition](https://mlsanity.com/paper/2004.03066/) (2020)
- [minicons: Enabling Flexible Behavioral and Representational Analyses of Transformer Language Models](https://mlsanity.com/paper/2203.13112/) (2022)
- [Cross-Linguistic Syntactic Evaluation of Word Prediction Models](https://mlsanity.com/paper/2005.00187/) (2020)
- [The EOS Decision and Length Extrapolation](https://mlsanity.com/paper/2010.07174/) (2020)
- [Structural Persistence in Language Models: Priming as a Window into Abstract Language Representations](https://mlsanity.com/paper/2109.14989/) (2021)
- [Exploring BERT's Sensitivity to Lexical Cues using Tests from Semantic Priming](https://mlsanity.com/paper/2010.03010/) (2020)

## Built on by (most cited first, 22 of 22)

- [Language Model Behavior: A Comprehensive Survey](https://mlsanity.com/paper/2303.11504/) (2023)
- [Evaluating Large Language Models on Controlled Generation Tasks](https://mlsanity.com/paper/2310.14542/) (2023)
- [Testing AI on language comprehension tasks reveals insensitivity to underlying meaning](https://mlsanity.com/paper/2302.12313/) (2023)
- [Changing Answer Order Can Decrease MMLU Accuracy](https://mlsanity.com/paper/2406.19470/) (2024)
- [ReCOGS: How Incidental Details of a Logical Form Overshadow an Evaluation of Semantic Interpretation](https://mlsanity.com/paper/2303.13716/) (2023)
- [Log Probabilities Are a Reliable Estimate of Semantic Plausibility in Base and Instruction-Tuned Language Models](https://mlsanity.com/paper/2403.14859/) (2024)
- [Do Language Models Exhibit Human-like Structural Priming Effects?](https://mlsanity.com/paper/2406.04847/) (2024)
- [Brittlebench: Quantifying LLM robustness via prompt sensitivity](https://mlsanity.com/paper/2603.13285/) (2026)
- [Robust Pronoun Fidelity with English LLMs: Are they Reasoning, Repeating, or Just Biased?](https://mlsanity.com/paper/2404.03134/) (2024)
- [What makes a good metric? Evaluating automatic metrics for text-to-image consistency](https://mlsanity.com/paper/2412.13989/) (2024)
- [Rapid Word Learning Through Meta In-Context Learning](https://mlsanity.com/paper/2502.14791/) (2025)
- [Experimental Contexts Can Facilitate Robust Semantic Property Inference in Language Models, but Inconsistently](https://mlsanity.com/paper/2401.06640/) (2024)
- [Who Relies More on World Knowledge and Bias for Syntactic Ambiguity Resolution: Humans or LLMs?](https://mlsanity.com/paper/2503.10838/) (2025)
- [Negative Pre-activations Differentiate Syntax](https://mlsanity.com/paper/2509.24198/) (2025)
- [The Grammar of Transformers: A Systematic Review of Interpretability Research on Syntactic Knowledge in Language Models](https://mlsanity.com/paper/2601.19926/) (2026)
- [Implicit Representations of Grammaticality in Language Models](https://mlsanity.com/paper/2605.05197/) (2026)
- [Revisiting Acceptability Judgements](https://mlsanity.com/paper/2305.14091/) (2023)
- [To Drop or Not to Drop? Predicting Argument Ellipsis Judgments: A Case Study in Japanese](https://mlsanity.com/paper/2404.11315/) (2024)
- [On the scaling relationship between cloze probabilities and language model next-token prediction](https://mlsanity.com/paper/2602.17848/) (2026)
- [Extending Minimal Pairs with Ordinal Surprisal Curves and Entropy Across Applied Domains](https://mlsanity.com/paper/2603.14400/) (2026)
- [GRUFF: LLM Pronoun Fidelity, Reasoning, and Biases in German](https://mlsanity.com/paper/2605.30214/) (2026)
- [Patterns of Priming in Production: Lexical, Semantic and Structural Alignment in Language Model Generation](https://mlsanity.com/paper/2609.04484/) (2026)

## Abstract (arXiv)

Targeted syntactic evaluations of language models ask whether models show stable preferences for syntactically acceptable content over minimal-pair unacceptable inputs. Most targeted syntactic evaluation datasets ask models to make these judgements with just a single context-free sentence as input. This does not match language models' training regime, in which input sentences are always highly contextualized by the surrounding corpus. This mismatch raises an important question: how robust are models' syntactic judgements in different contexts? In this paper, we investigate the stability of language models' performance on targeted syntactic evaluations as we vary properties of the input context: the length of the context, the types of syntactic phenomena it contains, and whether or not there are violations of grammaticality. We find that model judgements are generally robust when placed in randomly sampled linguistic contexts. However, they are substantially unstable for contexts containing syntactic structures matching those in the critical test content. Among all tested models (GPT-2 and five variants of OPT), we significantly improve models' judgements by providing contexts with matching syntactic structures, and conversely significantly worsen them using unacceptable contexts with matching but violated syntactic structures. This effect is amplified by the length of the context, except for unrelated inputs. We show that these changes in model performance are not explainable by simple features matching the context and the test inputs, such as lexical overlap and dependency overlap. This sensitivity to highly specific syntactic features of the context can only be explained by the models' implicit in-context learning abilities.