mlsanityData to 14 Oct 2026

How Language Model Hallucinations Can Snowball

arXiv 2305.13534 ↗First posted 22 May 2023cs.CLcited by 311

Not on the map. Why?

Cited by 50 in LLM reasoning, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

A major risk of using language models in practical applications is their tendency to hallucinate incorrect statements. Hallucinations are often attributed to knowledge gaps in LMs, but we hypothesize that in some cases, when justifying previously generated hallucinations, LMs output false claims that they can separately recognize as incorrect. We construct three question-answering datasets where ChatGPT and GPT-4 often state an incorrect answer and offer an explanation with at least one incorrect claim. Crucially, we find that ChatGPT and GPT-4 can identify 67% and 87% of their own mistakes, respectively. We refer to this phenomenon as hallucination snowballing: an LM over-commits to early mistakes, leading to more mistakes that it otherwise would not make.

Key facts. Cited by 311 papers in the map (data through 14 Oct 2026). Builds on 22 of its 31 references that are in the map. First posted on arXiv 22 May 2023.

Builds on

22 of 31 references.

References · 22

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  4. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  5. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
  6. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  7. 2019The Curious Case of Neural Text Degenerationcited by 2,916
  8. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  9. 2022Measuring and Narrowing the Compositionality Gap in Language Modelscited by 1,075
  10. 2020On Faithfulness and Factuality in Abstractive Summarizationcited by 900
  11. 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 886
  12. 2021Retrieval Augmentation Reduces Hallucination in Conversationcited by 692
10 more
  1. 2018Object Hallucination in Image Captioningcited by 669
  2. 2023Language Models can Solve Computer Taskscited by 372
  3. 2023Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedbackcited by 353
  4. 2023Evaluating Verifiability in Generative Search Enginescited by 303
  5. 2022The Parallelism Tradeoff: Limitations of Log-Precision Transformerscited by 193
  6. 2021The Curious Case of Hallucinations in Neural Machine Translationcited by 161
  7. 2020On Exposure Bias, Hallucination and Domain Shift in Neural Machine Translationcited by 127
  8. 2022Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generationcited by 108
  9. 2022(QA)2: Question Answering with Questionable Assumptionscited by 51
  10. 2021Which Linguist Invented the Lightbulb? Presupposition Verification for Question-Answeringcited by 41

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 67; 2024: 99; 2025: 80; 2026: 65.

311 papers.

Sort:

By area:

  1. 2023A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questionscited by 1,601
  2. 2023Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedbackcited by 759
  3. 2023Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Modelscited by 698
  4. 2023OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocationcited by 478
  5. 2023A Survey of Hallucination in Large Foundation Modelscited by 368
  6. 2023Chain-of-Verification Reduces Hallucination in Large Language Modelscited by 324
  7. 2023Analyzing and Mitigating Object Hallucination in Large Vision-Language Modelscited by 316
  8. 2023How is ChatGPT's behavior changing over time?cited by 270
  9. 2024MiniCheck: Efficient Fact-Checking of LLMs on Grounding Documentscited by 257
  10. 2023Challenges and Applications of Large Language Modelscited by 254
  11. 2023Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Taskscited by 246
  12. 2024When Can LLMs Actually Correct Their Own Mistakes? A Critical Survey of Self-Correction of LLMscited by 240
38 more
  1. 2023Automatically Correcting Large Language Models: Surveying the landscape of diverse self-correction strategiescited by 238
  2. 2023Fine-tuning Language Models for Factualitycited by 221
  3. 2025Why Language Models Hallucinatecited by 220
  4. 2023Survey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-Specificitycited by 206
  5. 2023MuSR: Testing the Limits of Chain-of-thought with Multistep Soft Reasoningcited by 201
  6. 2023Navigate through Enigmatic Labyrinth A Survey of Chain of Thought Reasoning: Advances, Frontiers and Futurecited by 184
  7. 2023Can LLM-Generated Misinformation Be Detected?cited by 177
  8. 2024The Landscape of Emerging AI Agent Architectures for Reasoning, Planning, and Tool Calling: A Surveycited by 168
  9. 2024GPT-4V(ision) is a Human-Aligned Evaluator for Text-to-3D Generationcited by 153
  10. 2023Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselvescited by 123
  11. 2023ReviewerGPT? An Exploratory Study on Using Large Language Models for Paper Reviewingcited by 122
  12. 2023Calibrated Language Models Must Hallucinatecited by 115
  13. 2024Making Reasoning Matter: Measuring and Improving Faithfulness of Chain-of-Thought Reasoningcited by 110
  14. 2024Beyond Functional Correctness: Exploring Hallucinations in LLM-Generated Codecited by 107
  15. 2024InstructRAG: Instructing Retrieval-Augmented Generation via Self-Synthesized Rationalescited by 101
  16. 2024A Comprehensive Survey of Hallucination in Large Language, Image, Video and Audio Foundation Modelscited by 100
  17. 2023Evaluation and Analysis of Hallucination in Large Vision-Language Modelscited by 99
  18. 2023Shepherd: A Critic for Language Model Generationcited by 97
  19. 2024Self-Introspective Decoding: Alleviating Hallucinations for Large Vision-Language Modelscited by 94
  20. 2023Can Knowledge Graphs Reduce Hallucinations in LLMs? : A Surveycited by 92
  21. 2023SAC3: Reliable Hallucination Detection in Black-Box Language Models via Semantic-aware Cross-check Consistencycited by 90
  22. 2024VideoHallucer: Evaluating Intrinsic and Extrinsic Hallucinations in Large Video-Language Modelscited by 90
  23. 2023Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinementcited by 89
  24. 2024The Dawn After the Dark: An Empirical Study on Factuality Hallucination in Large Language Modelscited by 88
  25. 2023Factcheck-Bench: Fine-Grained Evaluation Benchmark for Automatic Fact-checkerscited by 87
  26. 2023Cognitive Mirage: A Review of Hallucinations in Large Language Modelscited by 84
  27. 2024Self-playing Adversarial Language Game Enhances LLM Reasoningcited by 83
  28. 2023Exposing Attention Glitches with Flip-Flop Language Modelingcited by 75
  29. 2023Large Language Models can Learn Rulescited by 69
  30. 2023Attention Satisfies: A Constraint-Satisfaction Lens on Factual Errors of Language Modelscited by 67
  31. 2024Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systemscited by 66
  32. 2024Factuality of Large Language Models: A Surveycited by 66
  33. 2024Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedbackcited by 66
  34. 2024A Chain-of-Thought Is as Strong as Its Weakest Link: A Benchmark for Verifiers of Reasoning Chainscited by 65
  35. 2024Navigating the Risks: A Survey of Security, Privacy, and Ethics Threats in LLM-Based Agentscited by 65
  36. 2025The Illusion of Diminishing Returns: Measuring Long Horizon Execution in LLMscited by 64
  37. 2024Reasoning Aware Self-Consistency: Leveraging Reasoning Paths for Efficient LLM Samplingcited by 63
  38. 2024SPUQ: Perturbation-Based Uncertainty Quantification for Large Language Modelscited by 56

50 most cited of 311 All 311 (JSON)