mlsanityData to 7 Oct 2026

Are Language Models Worse than Humans at Following Prompts? It's Complicated

arXiv 2301.07085 ↗First posted 17 Jan 2023cs.CL cs.AIcited by 14

Not on the map. Why?

Abstract (arXiv)

Prompts have been the center of progress in advancing language models' zero-shot and few-shot performance. However, recent work finds that models can perform surprisingly well when given intentionally irrelevant or misleading prompts. Such results may be interpreted as evidence that model behavior is not "human like". In this study, we challenge a central assumption in such work: that humans would perform badly when given pathological instructions. We find that humans are able to reliably ignore irrelevant instructions and thus, like models, perform well on the underlying task despite an apparent lack of signal regarding the task they are being asked to do. However, when given deliberately misleading instructions, humans follow the instructions faithfully, whereas models do not. Our findings caution that future research should not idealize human behaviors as a monolith and should not train or evaluate models to mimic assumptions about these behaviors without first validating humans' behaviors empirically.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 28 of its 43 references that are in the map. First posted on arXiv 17 Jan 2023.

Builds on

28 of 43 references.

References · 28

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2022Training language models to follow instructions with human feedbackcited by 15,712
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  4. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  5. 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,769
  6. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,747
  7. 2022Scaling Instruction-Finetuned Language Modelscited by 2,997
  8. 2017A Broad-Coverage Challenge Corpus for Sentence Understanding through Inferencecited by 2,665
  9. 2015Character-level Convolutional Networks for Text Classificationcited by 2,293
  10. 2022Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language modelscited by 1,970
  11. 2021Multitask Prompted Training Enables Zero-Shot Task Generalizationcited by 1,462
  12. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?cited by 1,413
16 more
  1. 2022Least-to-Most Prompting Enables Complex Reasoning in Large Language Modelscited by 1,397
  2. 2018XNLI: Evaluating Cross-lingual Sentence Representationscited by 913
  3. 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 871
  4. 2019BLiMP: The Benchmark of Linguistic Minimal Pairs for Englishcited by 553
  5. 2016Solving General Arithmetic Word Problemscited by 353
  6. 2021Do Prompt-Based Models Really Understand the Meaning of their Prompts?cited by 324
  7. 2022Can language models learn from explanations in context?cited by 248
  8. 2022Language models show human-like content effects on reasoning taskscited by 171
  9. 2022The Unreliability of Explanations in Few-shot Prompting for Textual Reasoningcited by 169
  10. 2022Text and Patterns: For Effective Chain of Thought, It Takes Two to Tangocited by 128
  11. 2022GrIPS: Gradient-free, Edit-based Instruction Search for Prompting Large Language Modelscited by 127
  12. 2021Prompt Waywardness: The Curious Case of Discretized Interpretation of Continuous Promptscited by 64
  13. 2022Can Large Language Models Truly Understand Prompts? A Case Study with Negated Promptscited by 56
  14. 2021Beyond Fair Pay: Ethical Implications of NLP Crowdsourcingcited by 52
  15. 2022Can language models handle recursively nested grammatical structures? A case study on comparing models and humanscited by 41
  16. 2022Entailment Semantics Can Be Extracted from an Ideal Language Modelcited by 10

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 6; 2024: 6; 2025: 0; 2026: 2.

14 papers.

Sort:

By area:

  1. 2023Do LLMs exhibit human-like response biases? A case study in survey designcited by 176
  2. 2024PERSONA: A Reproducible Testbed for Pluralistic Alignmentcited by 71
  3. 2023LLMs as Workers in Human-Computational Algorithms? Replicating Crowdsourcing Pipelines with LLMscited by 27
  4. 2023Revisiting Prompt Engineering via Declarative Crowdsourcingcited by 12
  5. 2023Prompting a Large Language Model to Generate Diverse Motivational Messages: A Comparison with Human-Written Messagescited by 10
  6. 2023Expectations over Unspoken Alternatives Predict Pragmatic Inferencescited by 9
  7. 2024Conditional and Modal Reasoning in Large Language Modelscited by 9
  8. 2024Do Pre-Trained Language Models Detect and Understand Semantic Underspecification? Ask the DUST!cited by 7
  9. 2024Belief in the Machine: Investigating Epistemological Blind Spots of Language Modelscited by 5
  10. 2023Enhancing Programming eTextbooks with ChatGPT Generated Counterfactual-Thinking-Inspired Questionscited by 1
  11. 2024Do Language Models Understand the Cognitive Tasks Given to Them? Investigations with the N-Back Paradigmcited by 1
  12. 2026It's Not What You Say, It's How You Say It: Evaluating LLM Responses to Expressions of Beliefcited by 1
2 more
  1. 2024Can Language Model Understand Word Semantics as A Chatbot? An Empirical Study of Language Model Internal External Mismatchcited by 0
  2. 2026Exposure is not manifestation: measurement target and output resolution jointly determine which behavioural-faithfulness evaluator winscited by 0