mlsanityData to 7 Oct 2026

A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation

arXiv 2104.08704 ↗First posted 18 Apr 2021cs.CL cs.AIcited by 137

Not on the map. Why?

Abstract (arXiv)

Large pretrained generative models like GPT-3 often suffer from hallucinating non-existent or incorrect content, which undermines their potential merits in real applications. Existing work usually attempts to detect these hallucinations based on a corresponding oracle reference at a sentence or document level. However ground-truth references may not be readily available for many free-form text generation applications, and sentence- or document-level detection may fail to provide the fine-grained signals that would prevent fallacious content in real time. As a first step to addressing these issues, we propose a novel token-level, reference-free hallucination detection task and an associated annotated dataset named HaDes (HAllucination DEtection dataSet). To create this dataset, we first perturb a large number of text segments extracted from English language Wikipedia, and then verify these with crowd-sourced annotations. To mitigate label imbalance during annotation, we utilize an iterative model-in-loop strategy. We conduct comprehensive data analyses and create multiple baseline models.

Key facts. Cited by 137 papers in the map (data through 7 Oct 2026). Builds on 43 of its 64 references that are in the map. First posted on arXiv 18 Apr 2021.

Builds on

43 of 64 references.

References · 43

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2020Language Models are Few-Shot Learnerscited by 32,199
  4. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  5. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  6. 2019XLNet: Generalized Autoregressive Pretraining for Language Understandingcited by 2,801
  7. 2018FEVER: a large-scale dataset for Fact Extraction and VERificationcited by 1,362
  8. 2020On Faithfulness and Factuality in Abstractive Summarizationcited by 895
  9. 2019Adversarial NLI: A New Benchmark for Natural Language Understandingcited by 798
  10. 2018Object Hallucination in Image Captioningcited by 661
  11. 2019Evaluating the Factual Consistency of Abstractive Text Summarizationcited by 578
  12. 2017Adversarial Examples for Evaluating Reading Comprehension Systemscited by 568
31 more
  1. 2019Defending Against Neural Fake Newscited by 553
  2. 2020Asking and Answering Questions to Evaluate the Factual Consistency of Summariescited by 376
  3. 2018SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inferencecited by 333
  4. 2018Fast Abstractive Summarization with Reinforce-Selected Sentence Rewritingcited by 158
  5. 2020Detecting Hallucinated Content in Conditional Neural Sequence Generationcited by 151
  6. 2020On Exposure Bias, Hallucination and Domain Shift in Neural Machine Translationcited by 127
  7. 2019Handling Divergent Reference Texts when Evaluating Table-to-Text Generationcited by 119
  8. 2019Fake News Detection on Social Media using Geometric Deep Learningcited by 115
  9. 2019Assessing The Factual Accuracy of Generated Textcited by 103
  10. 2018Automated Fact Checking: Task formulations, methods and future directionscited by 96
  11. 2019Optimizing the Factual Correctness of a Summary: A Study of Summarizing Radiology Reportscited by 96
  12. 2019Answers Unite! Unsupervised Metrics for Reinforced Summarization Modelscited by 92
  13. 2019Do Massively Pretrained Language Models Make Better Storytellers?cited by 90
  14. 2020Shortformer: Better Language Modeling using Shorter Inputscited by 76
  15. 2019Step-by-Step: Separating Planning from Realization in Neural Data-to-Text Generationcited by 75
  16. 2020Multi-Fact Correction in Abstractive Text Summarizationcited by 66
  17. 2019Question Answering as an Automatic Evaluation Metric for News Article Summarizationcited by 63
  18. 2019Sticking to the Facts: Confident Decoding for Faithful Data-to-Text Generationcited by 62
  19. 2018Learning Neural Templates for Text Generationcited by 61
  20. 2020Towards Faithful Neural Table-to-Text Generation with Content-Matching Constraintscited by 48
  21. 2021Controlling Hallucinations at Word Level in Data-to-Text Generationcited by 41
  22. 2020Neural Deepfake Detection with Factual Structure of Textcited by 40
  23. 2019Constrained Decoding for Neural NLG from Compositional Representations in Task-Oriented Dialoguecited by 35
  24. 2020Neural Data-to-Text Generation via Jointly Learning the Segmentation and Correspondencecited by 25
  25. 2021Towards Faithfulness in Open Domain Table-to-text Generation from an Entity-centric Viewcited by 24
  26. 2019Sentence-Level Content Planning and Style Specification for Neural Text Generationcited by 23
  27. 2017Order-Planning Neural Text Generation From Structured Datacited by 20
  28. 2019Key Fact as Pivot: A Two-Stage Model for Low Resource Table-to-Text Generationcited by 20
  29. 2020Variational Template Machine for Data-to-Text Generationcited by 19
  30. 2020Posterior Control of Blackbox Generationcited by 16
  31. 2020Schema-Guided Natural Language Generationcited by 11

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 3; 2022: 6; 2023: 27; 2024: 59; 2025: 32; 2026: 10.

137 papers.

Sort:

By area:

  1. 2022Survey of Hallucination in Natural Language Generationcited by 2,056
  2. 2023SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Modelscited by 787
  3. 2023Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Modelscited by 695
  4. 2024LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinationscited by 230
  5. 2023FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Setscited by 198
  6. 2023Med-HALT: Medical Domain Hallucination Test for Large Language Modelscited by 156
  7. 2023Sociotechnical Safety Evaluation of Generative AI Systemscited by 154
  8. 2021CLIFF: Contrastive Learning for Improving Faithfulness and Factuality in Abstractive Summarizationcited by 120
  9. 2023Generating Benchmarks for Factuality Evaluation of Language Modelscited by 103
  10. 2024Large Language Models Meet NLP: A Surveycited by 102
  11. 2023Enhancing Uncertainty-Based Hallucination Detection with Stronger Focuscited by 98
  12. 2023ExpertQA: Expert-Curated Questions and Attributed Answerscited by 96
38 more
  1. 2023SAC3: Reliable Hallucination Detection in Black-Box Language Models via Semantic-aware Cross-check Consistencycited by 89
  2. 2025HalluLens: LLM Hallucination Benchmarkcited by 88
  3. 2024Fast Adversarial Attacks on Language Models In One GPU Minutecited by 85
  4. 2024FaithEval: Can Your Language Model Stay Faithful to Context, Even If "The Moon is Made of Marshmallows"cited by 85
  5. 2024Unsupervised Real-Time Hallucination Detection based on the Internal States of Large Language Modelscited by 84
  6. 2024Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedbackcited by 64
  7. 2024Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systemscited by 63
  8. 2024DRAGIN: Dynamic Retrieval Augmented Generation based on the Information Needs of Large Language Modelscited by 63
  9. 2023Red Teaming Language Model Detectors with Language Modelscited by 55
  10. 2023RECALL: A Benchmark for LLMs Robustness against External Counterfactual Knowledgecited by 51
  11. 2023Distilling Script Knowledge from Large Language Models for Constrained Language Planningcited by 45
  12. 2022Event knowledge in large language models: the gap between the impossible and the unlikelycited by 44
  13. 2023Decoder-Only or Encoder-Decoder? Interpreting Language Model as a Regularized Encoder-Decodercited by 44
  14. 2024Multicalibration for Confidence Scoring in LLMscited by 44
  15. 2024Hallucination Detection: Robustly Discerning Reliable Answers in Large Language Modelscited by 44
  16. 2023DelucionQA: Detecting Hallucinations in Domain-specific Question Answeringcited by 43
  17. 2024Estimating Knowledge in Large Language Models Without Generating a Single Tokencited by 39
  18. 2023Language Models Hallucinate, but May Excel at Fact Verificationcited by 35
  19. 2024AI Hallucinations: A Misnomer Worth Clarifyingcited by 34
  20. 2023Inference-Time Policy Adapters (IPA): Tailoring Extreme-Scale LMs without Fine-tuningcited by 33
  21. 2025MedHallu: A Comprehensive Benchmark for Detecting Medical Hallucinations in Large Language Modelscited by 33
  22. 2023Assessing Language Model Deployment with Risk Cardscited by 30
  23. 2025SemEval-2025 Task 3: Mu-SHROOM, the Multilingual Shared Task on Hallucinations and Related Observable Overgeneration Mistakescited by 27
  24. 2022RuCoLA: Russian Corpus of Linguistic Acceptabilitycited by 26
  25. 2024InterrogateLLM: Zero-Resource Hallucination Detection in LLM-Generated Answerscited by 26
  26. 2024SemEval-2024 Shared Task 6: SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakescited by 24
  27. 2021Learning to Complete Code with Sketchescited by 23
  28. 2022ThinkSum: Probabilistic reasoning over sets using large language modelscited by 23
  29. 2024Small Language Model Can Self-correctcited by 23
  30. 2021Am I Me or You? State-of-the-Art Dialogue Models Cannot Maintain an Identitycited by 22
  31. 2023UHGEval: Benchmarking the Hallucination of Chinese Large Language Models via Unconstrained Generationcited by 22
  32. 2024Benchmarking Large Multimodal Models against Common Corruptionscited by 22
  33. 2024RDRec: Rationale Distillation for LLM-based Recommendationcited by 21
  34. 2024A Probabilistic Framework for LLM Hallucination Detection via Belief Tree Propagationcited by 21
  35. 2024Hallucination Detection and Hallucination Mitigation: An Investigationcited by 20
  36. 2024MedHalu: Hallucinations in Responses to Healthcare Queries by Large Language Modelscited by 20
  37. 2023Zero-Resource Hallucination Prevention for Large Language Modelscited by 19
  38. 2024Stress-Testing Long-Context Language Models with Lifelong ICL and Task Haystackcited by 19

50 most cited of 137 All 137 (JSON)