mlsanityData to 14 Oct 2026

Understanding the Prompt Sensitivity

arXiv 2604.18389 ↗First posted 20 Apr 2026cs.CLcited by 3

Not on the map. Why?

Abstract (arXiv)

Prompt sensitivity, which refers to how strongly the output of a large language model (LLM) depends on the exact wording of its input prompt, raises concerns among users about the LLM's stability and reliability. In this work, we consider LLMs as multivariate functions and perform a first-order Taylor expansion, thereby analyzing the relationship between meaning-preserving prompts, their gradients, and the log probabilities of the model's next token. We derive an upper bound on the difference between log probabilities using the Cauchy-Schwarz inequality. We show that LLMs do not internally cluster similar inputs like smaller neural networks do, but instead disperse them. This dispersing behavior leads to an excessively high upper bound on the difference of log probabilities between two meaning-preserving prompts, making it difficult to effectively reduce to 0. In our analysis, we also show which types of meaning-preserving prompt variants are more likely to introduce prompt sensitivity risks in LLMs. In addition, we demonstrate that the upper bound is strongly correlated with an existing prompt sensitivity metric, PromptSensiScore. Moreover, by analyzing the logit variance, we find that prompt templates typically exert a greater influence on logits than the questions themselves. Overall, our results provide a general interpretation for why current LLMs can be highly sensitive to prompts with the same meaning, offering crucial evidence for understanding the prompt sensitivity of LLMs. Code for experiments is available at https://github.com/ku-nlp/Understanding_the_Prompt_Sensitivity.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 32 of its 65 references that are in the map. First posted on arXiv 20 Apr 2026.

Builds on

32 of 65 references.

References · 32

  1. 2017Attention Is All You Needcited by 60,361
  2. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  3. 2020Language Models are Few-Shot Learnerscited by 32,419
  4. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  5. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  6. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,886
  7. 2017Deep reinforcement learning from human preferencescited by 3,912
  8. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
  9. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  10. 2018Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answeringcited by 2,164
  11. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
  12. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?cited by 1,434
20 more
  1. 2021Calibrate Before Use: Improving Few-Shot Performance of Language Modelscited by 1,392
  2. 2020Making Pre-trained Language Models Better Few-shot Learnerscited by 1,256
  3. 2021What Makes Good In-Context Examples for GPT-3?cited by 1,195
  4. 2021Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivitycited by 1,135
  5. 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 817
  6. 2019Are Transformers universal approximators of sequence-to-sequence functions?cited by 406
  7. 2023State of What Art? A Call for Multi-Prompt LLM Evaluationcited by 310
  8. 2016Large-Margin Softmax Loss for Convolutional Neural Networkscited by 296
  9. 2022Selective Annotation Makes Language Models Better Few-Shot Learnerscited by 243
  10. 2023How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Compositioncited by 223
  11. 2015Visualizing and Understanding Neural Models in NLPcited by 200
  12. 2024ProSA: Assessing and Understanding the Prompt Sensitivity of LLMscited by 149
  13. 2022Interpreting Language Models with Contrastive Explanationscited by 92
  14. 2024Mind Your Format: Towards Consistent Evaluation of In-Context Learning Improvementscited by 83
  15. 2024POSIX: A Prompt Sensitivity Index For Large Language Modelscited by 59
  16. 2022Robustness of Learning from Task Instructionscited by 35
  17. 2025On the Alignment of Large Language Models with Global Human Opinioncited by 13
  18. 2024Take the essence and discard the dross: A Rethinking on Data Selection for Fine-Tuning Large Language Modelscited by 12
  19. 2019G-softmax: Improving Intra-class Compactness and Inter-class Separability of Featurescited by 6
  20. 2024Transformer-based Causal Language Models Perform Clusteringcited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Discourse-Role Labels as Presentation-Time Variables for Context Use in Language Modelscited by 1
  2. 2026How Much Do LLM-as-a-Judge Design Choices Matter? A Systematic Comparison of Prompt Designs, Rating Scales, and Modelscited by 0
  3. 2026A Distributional Framework for Behavioral Characterization and Risk Management of Large Language Modelscited by 0