mlsanityData to 7 Oct 2026

Mind the instructions: a holistic evaluation of consistency and interactions in prompt-based learning

arXiv 2310.13486 ↗First posted 20 Oct 2023cs.CL cs.AIcited by 31

Not on the map. Why?

Abstract (arXiv)

Finding the best way of adapting pre-trained language models to a task is a big challenge in current NLP. Just like the previous generation of task-tuned models (TT), models that are adapted to tasks via in-context-learning (ICL) are robust in some setups but not in others. Here, we present a detailed analysis of which design choices cause instabilities and inconsistencies in LLM predictions. First, we show how spurious correlations between input distributions and labels -- a known issue in TT models -- form only a minor problem for prompted models. Then, we engage in a systematic, holistic evaluation of different factors that have been found to influence predictions in a prompting setup. We test all possible combinations of a range of factors on both vanilla and instruction-tuned (IT) LLMs of different scale and statistically analyse the results to show which factors are the most influential, interactive or stable. Our results show which factors can be used without precautions and which should be avoided or handled with care in most settings.

Key facts. Cited by 31 papers in the map (data through 7 Oct 2026). Builds on 42 of its 57 references that are in the map. First posted on arXiv 20 Oct 2023.

Builds on

42 of 57 references.

References · 42

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2022Training language models to follow instructions with human feedbackcited by 15,712
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  4. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  5. 2016SQuAD: 100,000+ Questions for Machine Comprehension of Textcited by 4,115
  6. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,747
  7. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,674
  8. 2017A Broad-Coverage Challenge Corpus for Sentence Understanding through Inferencecited by 2,665
  9. 2019HuggingFace's Transformers: State-of-the-art Natural Language Processingcited by 2,201
  10. 2020Shortcut Learning in Deep Neural Networkscited by 1,928
  11. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?cited by 1,413
  12. 2021Calibrate Before Use: Improving Few-Shot Performance of Language Modelscited by 1,371
30 more
  1. 2021Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivitycited by 1,122
  2. 2023LIMA: Less Is More for Alignmentcited by 1,080
  3. 2023A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivitycited by 888
  4. 2022LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scalecited by 881
  5. 2019Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inferencecited by 864
  6. 2019Adversarial NLI: A New Benchmark for Natural Language Understandingcited by 798
  7. 2022What Can Transformers Learn In-Context? A Case Study of Simple Function Classescited by 700
  8. 2022What learning algorithm is in-context learning? Investigations with linear modelscited by 671
  9. 2018Annotation Artifacts in Natural Language Inference Datacited by 656
  10. 2017Adversarial Examples for Evaluating Reading Comprehension Systemscited by 568
  11. 2019Learning the Difference that Makes a Difference with Counterfactually-Augmented Datacited by 428
  12. 2019PAWS: Paraphrase Adversaries from Word Scramblingcited by 395
  13. 2023Larger language models do in-context learning differentlycited by 386
  14. 2019Using Pre-Training Can Improve Model Robustness and Uncertaintycited by 373
  15. 2020Pretrained Transformers Improve Out-of-Distribution Robustnesscited by 293
  16. 2018Hypothesis Only Baselines in Natural Language Inferencecited by 291
  17. 2022PromptSource: An Integrated Development Environment and Repository for Natural Language Promptscited by 288
  18. 2022Prompting GPT-3 To Be Reliablecited by 258
  19. 2019Probing Neural Network Comprehension of Natural Language Argumentscited by 212
  20. 2021Reframing Instructional Prompts to GPTk's Languagecited by 176
  21. 2019Are We Modeling the Task or the Annotator? An Investigation of Annotator Bias in Natural Language Understanding Datasetscited by 173
  22. 2023Few-shot Fine-tuning vs. In-context Learning: A Fair Comparison and Evaluationcited by 134
  23. 2020Beat the AI: Investigating Adversarial Human Annotation for Reading Comprehensioncited by 131
  24. 2021Adapting Language Models for Zero-shot Learning by Meta-tuning on Dataset and Prompt Collectionscited by 128
  25. 2017Bilateral Multi-Perspective Matching for Natural Language Sentencescited by 120
  26. 2022Ground-Truth Labels Matter: A Deeper Look into Input-Label Demonstrationscited by 116
  27. 2020The Effect of Natural Distribution Shift on Question Answering Modelscited by 104
  28. 2022State-of-the-art generalisation research in NLP: A taxonomy and reviewcited by 97
  29. 2021Prompt Waywardness: The Curious Case of Discretized Interpretation of Continuous Promptscited by 64
  30. 2019When Choosing Plausible Alternatives, Clever Hans can be Clevercited by 39

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 4; 2024: 17; 2025: 7; 2026: 3.

31 papers.

Sort:

By area:

  1. 2023State of What Art? A Call for Multi-Prompt LLM Evaluationcited by 299
  2. 2024tinyBenchmarks: evaluating LLMs with fewer examplescited by 295
  3. 2024When Benchmarks are Targets: Revealing the Sensitivity of Large Language Model Leaderboardscited by 134
  4. 2024Is Temperature the Creativity Parameter of Large Language Models?cited by 92
  5. 2024Efficient multi-prompt evaluation of LLMscited by 72
  6. 2024Quantifying Variance in Evaluation Benchmarkscited by 52
  7. 2024Examining the robustness of LLM evaluation to the distributional assumptions of benchmarkscited by 31
  8. 2024MERA: A Comprehensive LLM Evaluation in Russiancited by 22
  9. 2025MultiLoKo: a multilingual local knowledge benchmark for LLMs spanning 31 languagescited by 22
  10. 2024LiSeCo: Linear Semantic Control for Language Generationcited by 20
  11. 2023WorldSense: A Synthetic Benchmark for Grounded Reasoning in Large Language Modelscited by 19
  12. 2023Data Management For Training Large Language Models: A Surveycited by 17
19 more
  1. 2024Comparing Specialised Small and General Large Language Models on Text Classification: 100 Labelled Samples to Achieve Break-Even Performancecited by 17
  2. 2024Towards Rationality in Language and Multimodal Agents: A Surveycited by 15
  3. 2023A Survey on Stability of Learning with Limited Labelled Data and its Sensitivity to the Effects of Randomnesscited by 14
  4. 2024SelfPrompt: Autonomously Evaluating LLM Robustness via Domain-Constrained Knowledge Guidelines and Refined Adversarial Promptscited by 10
  5. 2024Searching for Structure: Investigating Emergent Communication with Large Language Modelscited by 10
  6. 2025On the Consistency of Multilingual Context Utilization in Retrieval-Augmented Generationcited by 10
  7. 2024From Form(s) to Meaning: Probing the Semantic Depths of Language Models Using Multisense Consistencycited by 7
  8. 2025A Modular Taxonomy for Hate Speech Definitions and Its Impact on Zero-Shot LLM Classification Performancecited by 6
  9. 2025Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustnesscited by 5
  10. 2024KGPA: Robustness Evaluation for Large Language Models via Cross-Domain Knowledge Graphscited by 3
  11. 2024TaskEval: Assessing Difficulty of Code Generation Tasks for Large Language Modelscited by 3
  12. 2025Can LLMs Help Uncover Insights about LLMs? A Large-Scale, Evolving Literature Analysis of Frontier LLMscited by 3
  13. 2026Better as Generators Than Classifiers: Leveraging LLMs and Synthetic Data for Low-Resource Multilingual Classificationcited by 2
  14. 2026LPDS: Evaluating LLM Robustness Through Logic-Preserving Difficulty Scalingcited by 2
  15. 2024On Sensitivity of Learning with Limited Labelled Data to the Effects of Randomness: Impact of Interactions and Systematic Choicescited by 1
  16. 2024Sample Design Engineering: An Empirical Study of What Makes Good Downstream Fine-Tuning Samples for LLMscited by 1
  17. 2025LASER: Stratified Selective Sampling for Instruction Tuning with Dedicated Scoring Strategycited by 1
  18. 2025Instance-level Randomization: Toward More Stable LLM Evaluationscited by 1
  19. 2026ChemDIRT: A Diversified Instruction, Representation, and Task Benchmark for Robust Chemistry-LLM Evaluationcited by 0