mlsanityData to 14 Oct 2026

Why Language Models Hallucinate

arXiv 2509.04664 ↗First posted 4 Sep 2025cs.CLcited by 220

Not on the map. Why?

Cited by 54 in LLM reasoning, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Like students facing hard exam questions, large language models sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty. Such "hallucinations" persist even in state-of-the-art systems and undermine trust. We argue that language models hallucinate because the training and evaluation procedures reward guessing over acknowledging uncertainty, and we analyze the statistical causes of hallucinations in the modern training pipeline. Hallucinations need not be mysterious -- they originate simply as errors in binary classification. If incorrect statements cannot be distinguished from facts, then hallucinations in pretrained language models will arise through natural statistical pressures. We then argue that hallucinations persist due to the way most evaluations are graded -- language models are optimized to be good test-takers, and guessing when uncertain improves test performance. This "epidemic" of penalizing uncertain responses can only be addressed through a socio-technical mitigation: modifying the scoring of existing benchmarks that are misaligned but dominate leaderboards, rather than introducing additional hallucination evaluations. This change may steer the field toward more trustworthy AI systems.

Key facts. Cited by 220 papers in the map (data through 14 Oct 2026). Builds on 44 of its 82 references that are in the map. First posted on arXiv 4 Sep 2025.

Builds on

44 of 82 references.

References · 44

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  5. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,373
  6. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,050
  7. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  8. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  9. 2022Language Models (Mostly) Know What They Knowcited by 1,856
  10. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  11. 2022Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Themcited by 1,807
  12. 2021WebGPT: Browser-assisted question-answering with human feedbackcited by 1,745
32 more
  1. 2020On Faithfulness and Factuality in Abstractive Summarizationcited by 900
  2. 2015Scheduled Sampling for Sequence Prediction with Recurrent Neural Networkscited by 803
  3. 2022Teaching Models to Express Their Uncertainty in Wordscited by 751
  4. 2021BBQ: A Hand-Built Bias Benchmark for Question Answeringcited by 701
  5. 2021Retrieval Augmentation Reduces Hallucination in Conversationcited by 692
  6. 2023The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"cited by 407
  7. 2023How Language Model Hallucinations Can Snowballcited by 311
  8. 2023Do Large Language Models Know What They Don't Know?cited by 243
  9. 2024Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?cited by 232
  10. 2023Fine-tuning Language Models for Factualitycited by 221
  11. 2024WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wildcited by 218
  12. 2020Reducing conversational agents' overconfidence through linguistic calibrationcited by 210
  13. 2022Factuality Enhanced Language Models for Open-Ended Text Generationcited by 202
  14. 2023MuSR: Testing the Limits of Chain-of-thought with Multistep Soft Reasoningcited by 201
  15. 2023Head-to-Tail: How Knowledgeable are Large Language Models (LLMs)? A.K.A. Will LLMs Replace Knowledge Graphs?cited by 168
  16. 2025Beyond Binary Rewards: Training LMs to Reason About Their Uncertaintycited by 117
  17. 2023Do Language Models Know When They're Hallucinating References?cited by 116
  18. 2023Calibrated Language Models Must Hallucinatecited by 115
  19. 2023NPHardEval: Dynamic Benchmark on Reasoning Ability of Large Language Models via Complexity Classescited by 79
  20. 2024Open-LLM-Leaderboard: From Multi-choice to Open-style Questions for LLMs Evaluation, Benchmark, and Arenacited by 76
  21. 2024GraphArena: Evaluating and Exploring Large Language Models on Graph Computationcited by 49
  22. 2022How Pre-trained Language Models Capture Factual Knowledge? A Causal-Inspired Analysiscited by 42
  23. 2024Language Generation in the Limitcited by 40
  24. 2025Pragmatics in the Era of Large Language Models: A Survey on Datasets, Evaluation, Opportunities and Challengescited by 31
  25. 2025Verify when Uncertain: Beyond Self-Consistency in Black Box Hallucination Detectioncited by 18
  26. 2021Investigating Memorization of Conspiracy Theories in Text Generationcited by 15
  27. 2025Why and How LLMs Hallucinate: Connecting the Dots with Subsequence Associationscited by 15
  28. 2018Actively Avoiding Nonsense in Generative Modelscited by 12
  29. 2023Hijacking Context in Large Multi-modal Modelscited by 6
  30. 2025Answer, Refuse, or Guess? Investigating Risk-Aware Decision Making in Language Modelscited by 6
  31. 2015Learning Reductions that Really Workcited by 3
  32. 2025Hallucination, Monofacts, and Miscalibration: An Empirical Investigationcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 66; 2026: 154.

220 papers.

Sort:

By area:

  1. 2025TruthRL: Incentivizing Truthful LLMs via Reinforcement Learningcited by 27
  2. 2025From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Modelscited by 26
  3. 2025Training LLMs for Honesty via Confessionscited by 22
  4. 2025Probing the Critical Point (CritPt) of AI Reasoning: a Frontier Physics Research Benchmarkcited by 21
  5. 2026Spherical Steering: Geometry-Aware Activation Rotation for Language Modelscited by 21
  6. 2025Trace Length is a Simple Uncertainty Signal in Reasoning Modelscited by 19
  7. 2025Mitigating LLM Hallucination via Behaviorally Calibrated Reinforcement Learningcited by 18
  8. 2025RAG Meets Temporal Graphs: Time-Sensitive Modeling and Retrieval for Evolving Knowledgecited by 13
  9. 2026Over-Searching in Search-Augmented Large Language Modelscited by 11
  10. 2026Evaluating LLMs in Finance Requires Explicit Bias Considerationcited by 11
  11. 2025Failure Modes in LLM Systems: A System-Level Taxonomy for Reliable AI Applicationscited by 10
  12. 2026Causal Evidence that Language Models use Confidence to Drive Behaviorcited by 10
38 more
  1. 2025Language Generation: Complexity Barriers and Implications for Learningcited by 9
  2. 2026Hallucination Begins Where Saliency Dropscited by 9
  3. 2026The Illusion of Certainty: Decoupling Capability and Calibration in On-Policy Distillationcited by 9
  4. 2025MedMMV: A Controllable Multimodal Multi-Agent Framework for Reliable and Verifiable Clinical Reasoningcited by 8
  5. 2026ThinkRouter: Efficient Reasoning via Routing Thinking between Latent and Discrete Spacescited by 8
  6. 2025Mixture-of-Minds: Multi-Agent Reinforcement Learning for Table Understandingcited by 7
  7. 2025FAPO: Flawed-Aware Policy Optimization for Efficient and Reliable Reasoningcited by 7
  8. 2025TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Frameworkcited by 7
  9. 2026When Do LLM Preferences Predict Downstream Behavior?cited by 7
  10. 2026From Conflict to Consensus: Boosting Medical Reasoning via Multi-Round Agentic RAGcited by 7
  11. 2026BAS: A Decision-Theoretic Approach to Evaluating Large Language Model Confidencecited by 7
  12. 2026BayesBench: Evaluating LLM Belief Trajectories Under Multi-Turn Evidence Accumulationcited by 7
  13. 2025PsychiatryBench: A Multi-Task Benchmark for LLMs in Psychiatrycited by 6
  14. 2025Hallucination reduction with CASAL: Contrastive Activation Steering For Amortized Learningcited by 6
  15. 2025AdaSearch: Balancing Parametric Knowledge and Search in Large Language Models via Reinforcement Learningcited by 6
  16. 2026Reaching Beyond the Mode: RL for Distributional Reasoning in Language Modelscited by 6
  17. 2025ChartHal: A Fine-grained Framework Evaluating Hallucination of Large Vision Language Models in Chart Understandingcited by 5
  18. 2025Cognitive Load Limits in Large Language Models: Benchmarking Multi-Hop Reasoningcited by 5
  19. 2025Learning to Reason for Hallucination Span Detectioncited by 5
  20. 2025BrowseConf: Confidence-Guided Test-Time Scaling for Web Agentscited by 5
  21. 2025Honesty over Accuracy: Trustworthy Language Models through Reinforced Hesitationcited by 5
  22. 2026Are LLM Evaluators Really Narcissists? Sanity Checking Self-Preference Evaluationscited by 5
  23. 2026Characterizing, Evaluating, and Optimizing Complex Reasoningcited by 5
  24. 2026IslamicLegalBench: Evaluating LLMs Knowledge and Reasoning of Islamic Law Across 1,200 Years of Islamic Pluralist Legal Traditionscited by 5
  25. 2026MARCH: Multi-Agent Reinforced Self-Check for LLM Hallucinationcited by 5
  26. 2026Self-Improvement of Large Language Models: A Technical Overview and Future Outlookcited by 5
  27. 2026I-CALM: Incentivizing Confidence-Aware Abstention for LLM Selective Answeringcited by 5
  28. 2026Calibration-Aware Policy Optimization for Reasoning LLMscited by 5
  29. 2025XML Prompting as Grammar-Constrained Interaction: Fixed-Point Semantics, Convergence Guarantees, and Human-AI Protocolscited by 4
  30. 2025Generalization of Fine-Tuned Uncertainty Communication and Metacognition in Large Language Modelscited by 4
  31. 2025The Future of Generative AI in Software Engineering: A Vision from Industry and Academia in the European GENIUS Projectcited by 4
  32. 2025TriDF: Evaluating Perception, Detection, and Hallucination for Interpretable DeepFake Detectioncited by 4
  33. 2026LLMs Can't Play Hangman: On the Necessity of a Private Working Memory for Language Agentscited by 4
  34. 2026Feedback by Design: Understanding and Overcoming User Feedback Barriers in Conversational Agentscited by 4
  35. 2026FaithRL: Learning to Reason Faithfully through Step-Level Faithfulness Maximizationcited by 4
  36. 2026Stop Rewarding Hallucinated Steps: Faithfulness-Aware Step-Level Reinforcement Learning for Small Reasoning Modelscited by 4
  37. 2026Revis: Sparse Latent Steering to Mitigate Object Hallucination in Large Vision-Language Modelscited by 4
  38. 2026LLMs Exhibit Significantly Lower Uncertainty in Creative Writing Than Professional Writerscited by 4

50 most cited of 220 All 220 (JSON)