mlsanityData to 14 Oct 2026

In-Context Learning May Not Elicit Trustworthy Reasoning: A-Not-B Errors in Pretrained Language Models

arXiv 2409.15454 ↗First posted 23 Sep 2024cs.CL cs.AIcited by 8

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Recent advancements in artificial intelligence have led to the creation of highly capable large language models (LLMs) that can perform tasks in a human-like manner. However, LLMs exhibit only infant-level cognitive abilities in certain areas. One such area is the A-Not-B error, a phenomenon seen in infants where they repeat a previously rewarded behavior despite well-observed changed conditions. This highlights their lack of inhibitory control -- the ability to stop a habitual or impulsive response. In our work, we design a text-based multi-choice QA scenario similar to the A-Not-B experimental settings to systematically test the inhibitory control abilities of LLMs. We found that state-of-the-art LLMs (like Llama3-8b) perform consistently well with in-context learning (ICL) but make errors and show a significant drop of as many as 83.3% in reasoning tasks when the context changes trivially. This suggests that LLMs only have inhibitory control abilities on par with human infants in this regard, often failing to suppress the previously established response pattern during ICL.

Key facts. Cited by 8 papers in the map (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 58 of its 123 references that are in the map. First posted on arXiv 23 Sep 2024.

Builds on

58 of 123 references.

Closest ancestors · 5

Same direction, newest first

  1. 2024A Peek into Token Bias: Large Language Models Are Not Yet Genuine Reasoners16 Jun
  2. 2024Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking14 Mar
  3. 2023Large Language Models are In-context Teachers for Knowledge Reasoning12 Nov
  4. 2023Exploring the psychology of LLMs' Moral and Legal Reasoning2 Aug
  5. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?25 Feb

Other references · 53

  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. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  6. 2020Scaling Laws for Neural Language Modelscited by 6,586
  7. 2018GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understandingcited by 4,899
  8. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,886
  9. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,182
  10. 2023Generative Agents: Interactive Simulacra of Human Behaviorcited by 3,568
  11. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  12. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
38 more
  1. 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,651
  2. 2023Jailbreaking Black Box Large Language Models in Twenty Queriescited by 1,420
  3. 2022Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agentscited by 1,035
  4. 2022In-context Learning and Induction Headscited by 964
  5. 2021An Explanation of In-context Learning as Implicit Bayesian Inferencecited by 938
  6. 2014Deep Learning in Neural Networks: An Overviewcited by 770
  7. 2017Crowdsourcing Multiple Choice Science Questionscited by 765
  8. 2022A Survey on In-context Learningcited by 721
  9. 2019MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalismscited by 662
  10. 2024Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Surveycited by 656
  11. 2023Large Language Models Are Not Robust Multiple Choice Selectorscited by 486
  12. 2024GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projectioncited by 463
  13. 2023CMMLU: Measuring massive multitask language understanding in Chinesecited by 449
  14. 2023Cognitive Architectures for Language Agentscited by 429
  15. 2023LeanDojo: Theorem Proving with Retrieval-Augmented Language Modelscited by 427
  16. 2023The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"cited by 407
  17. 2024Hallucination is Inevitable: An Innate Limitation of Large Language Modelscited by 332
  18. 2023Large Language Models as Tool Makerscited by 303
  19. 2023Large Language Model Guided Tree-of-Thoughtcited by 221
  20. 2022Large Language Models with Controllable Working Memorycited by 169
  21. 2023Large Language Models Are Latent Variable Models: Explaining and Finding Good Demonstrations for In-Context Learningcited by 149
  22. 2023Decomposing Uncertainty for Large Language Models through Input Clarification Ensemblingcited by 114
  23. 2019Learning to Prove Theorems via Interacting with Proof Assistantscited by 109
  24. 2024Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Modelscited by 91
  25. 2023Attack Prompt Generation for Red Teaming and Defending Large Language Modelscited by 76
  26. 2023Evaluating Language Models for Mathematics through Interactionscited by 70
  27. 2023Large Language Models Can be Lazy Learners: Analyze Shortcuts in In-Context Learningcited by 69
  28. 2024Can multiple-choice questions really be useful in detecting the abilities of LLMs?cited by 69
  29. 2024Alice in Wonderland: Simple Tasks Showing Complete Reasoning Breakdown in State-Of-the-Art Large Language Modelscited by 68
  30. 2024LLMs achieve adult human performance on higher-order theory of mind taskscited by 61
  31. 2019SpatialSense: An Adversarially Crowdsourced Benchmark for Spatial Relation Recognitioncited by 55
  32. 2023Two Failures of Self-Consistency in the Multi-Step Reasoning of LLMscited by 36
  33. 2024Multiple-Choice Questions are Efficient and Robust LLM Evaluatorscited by 28
  34. 2024Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Taskscited by 24
  35. 2021Neuro-Symbolic AI: An Emerging Class of AI Workloads and their Characterizationcited by 22
  36. 2024ChatGPT Based Data Augmentation for Improved Parameter-Efficient Debiasing of LLMscited by 18
  37. 2022PromptAttack: Prompt-based Attack for Language Models via Gradient Searchcited by 15
  38. 2023Exploring Social Bias in Downstream Applications of Text-to-Image Foundation Modelscited by 7

50 most cited of 53 All 53 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 5; 2026: 3.

8 papers.

  1. 2025The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMscited by 36
  2. 2025Are Smarter LLMs Safer? Exploring Safety-Reasoning Trade-offs in Prompting and Fine-Tuningcited by 20
  3. 2025FusionFactory: Fusing LLM Capabilities with Multi-LLM Log Datacited by 10
  4. 2026Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMscited by 7
  5. 2026Interactive Evaluation Requires a Design Sciencecited by 3
  6. 2025seqBench: A Tunable Benchmark to Quantify Sequential Reasoning Limits of LLMscited by 2
  7. 2026SocialVeil: Probing Social Intelligence of Language Agents under Communication Barrierscited by 2
  8. 2025ICL Optimized Fragilitycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025A Review of Developmental Interpretability in Large Language Models
  2. 2023Comparing Machines and Children: Using Developmental Psychology Experiments to Assess the Strengths and Weaknesses of LaMDA Responses
  3. 2026BabyReasoningBench: Generating Developmentally-Inspired Reasoning Tasks for Evaluating Baby Language Models
  4. 2026The Age of Curiosity Meets the Age of AI: Benchmarking Child Safety in Large Language Models
  5. 2023Boosting Theory-of-Mind Performance in Large Language Models via Prompting
  6. 2023Evaluating Large Language Models in Theory of Mind Tasks
  7. 2026Evaluating Theory of Mind in Reasoning Models: Robustness over Reasoning
  8. 2025Short-Path Prompting in LLMs: Analyzing Reasoning Instability and Solutions for Robust Performance