mlsanityData to 14 Oct 2026

How Should Pre-Trained Language Models Be Fine-Tuned Towards Adversarial Robustness?

arXiv 2112.11668 ↗First posted 22 Dec 2021cs.CL cs.LGcited by 49

Not on the map. Why?

Abstract (arXiv)

The fine-tuning of pre-trained language models has a great success in many NLP fields. Yet, it is strikingly vulnerable to adversarial examples, e.g., word substitution attacks using only synonyms can easily fool a BERT-based sentiment analysis model. In this paper, we demonstrate that adversarial training, the prevalent defense technique, does not directly fit a conventional fine-tuning scenario, because it suffers severely from catastrophic forgetting: failing to retain the generic and robust linguistic features that have already been captured by the pre-trained model. In this light, we propose Robust Informative Fine-Tuning (RIFT), a novel adversarial fine-tuning method from an information-theoretical perspective. In particular, RIFT encourages an objective model to retain the features learned from the pre-trained model throughout the entire fine-tuning process, whereas a conventional one only uses the pre-trained weights for initialization. Experimental results show that RIFT consistently outperforms the state-of-the-arts on two popular NLP tasks: sentiment analysis and natural language inference, under different attacks across various pre-trained language models.

Key facts. Cited by 49 papers in the map (data through 14 Oct 2026). Builds on 65 of its 85 references that are in the map. First posted on arXiv 22 Dec 2021.

Builds on

65 of 85 references.

References · 65

  1. 2017Attention Is All You Needcited by 60,361
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  3. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
  4. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  5. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  6. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  7. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  8. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  9. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  10. 2013Intriguing properties of neural networkscited by 5,155
  11. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  12. 2013Auto-Encoding Variational Bayescited by 4,053
38 more
  1. 2013Distributed Representations of Words and Phrases and their Compositionalitycited by 3,474
  2. 2016Towards Evaluating the Robustness of Neural Networkscited by 2,998
  3. 2019XLNet: Generalized Autoregressive Pretraining for Language Understandingcited by 2,811
  4. 2016Learning without Forgettingcited by 2,496
  5. 2016iCaRL: Incremental Classifier and Representation Learningcited by 2,333
  6. 2018Deep contextualized word representationscited by 2,290
  7. 2017Gradient Episodic Memory for Continual Learningcited by 1,898
  8. 2015A large annotated corpus for learning natural language inferencecited by 1,847
  9. 2016Adversarial examples in the physical worldcited by 1,769
  10. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,517
  11. 2020Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hyperspherecited by 1,405
  12. 2018Learning deep representations by mutual information estimation and maximizationcited by 1,273
  13. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,268
  14. 2019BERT Rediscovers the Classical NLP Pipelinecited by 1,059
  15. 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 953
  16. 2016Adversarial Machine Learning at Scalecited by 936
  17. 2018Robustness May Be at Odds with Accuracycited by 880
  18. 2020What Makes for Good Views for Contrastive Learning?cited by 846
  19. 2015The Limitations of Deep Learning in Adversarial Settingscited by 790
  20. 2019Contrastive Representation Distillationcited by 748
  21. 2013An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networkscited by 722
  22. 2017Adversarial Examples for Evaluating Reading Comprehension Systemscited by 571
  23. 2019Gradient based sample selection for online continual learningcited by 571
  24. 2019On Variational Bounds of Mutual Informationcited by 559
  25. 2020On Adaptive Attacks to Adversarial Example Defensescited by 555
  26. 2020Overfitting in adversarially robust deep learningcited by 506
  27. 2018Generating Natural Language Adversarial Examplescited by 424
  28. 2018Black-box Generation of Adversarial Text Sequences to Evade Deep Learning Classifierscited by 345
  29. 2019On Mutual Information Maximization for Representation Learningcited by 316
  30. 2017Synthetic and Natural Noise Both Break Neural Machine Translationcited by 306
  31. 2018Adversarial Example Generation with Syntactically Controlled Paraphrase Networkscited by 286
  32. 2016Adversarial Training Methods for Semi-Supervised Text Classificationcited by 280
  33. 2020Revisiting Few-sample BERT Fine-tuningcited by 238
  34. 2019Visualizing and Measuring the Geometry of BERTcited by 216
  35. 2017Generating Natural Adversarial Examplescited by 183
  36. 2018Disentangling Adversarial Robustness and Generalizationcited by 172
  37. 2019Certified Robustness to Adversarial Word Substitutionscited by 168
  38. 2019Combating Adversarial Misspellings with Robust Word Recognitioncited by 163

50 most cited of 65 All 65 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 7; 2023: 16; 2024: 10; 2025: 12; 2026: 4.

49 papers.

Sort:

By area:

  1. 2023A Survey on Large Language Model (LLM) Security and Privacy: The Good, the Bad, and the Uglycited by 387
  2. 2023Investigating the Catastrophic Forgetting in Multimodal Large Language Modelscited by 119
  3. 2023Prompt as Triggers for Backdoor Attack: Examining the Vulnerability in Language Modelscited by 94
  4. 2023A Comprehensive Study on Robustness of Image Classification Models: Benchmarking and Rethinkingcited by 85
  5. 2023FLIRT: Feedback Loop In-context Red Teamingcited by 83
  6. 2023A Comprehensive Survey of Forgetting in Deep Learning Beyond Continual Learningcited by 69
  7. 2024Model Tailor: Mitigating Catastrophic Forgetting in Multi-modal Large Language Modelscited by 54
  8. 2024A Survey of Recent Backdoor Attacks and Defenses in Large Language Modelscited by 36
  9. 2023Preserving Pre-trained Features Helps Calibrate Fine-tuned Language Modelscited by 29
  10. 2024Unlearning Backdoor Attacks for LLMs with Weak-to-Strong Knowledge Distillationcited by 27
  11. 2022Certified Robustness Against Natural Language Attacks by Causal Interventioncited by 24
  12. 2023InfoCTM: A Mutual Information Maximization Perspective of Cross-Lingual Topic Modelingcited by 24
37 more
  1. 2022Robustness Analysis of Video-Language Models Against Visual and Language Perturbationscited by 23
  2. 2022Textual Manifold-based Defense Against Natural Language Adversarial Examplescited by 19
  3. 2023TWINS: A Fine-Tuning Framework for Improved Transferability of Adversarial Robustness and Generalizationcited by 18
  4. 2022Pre-trained Adversarial Perturbationscited by 16
  5. 2024Defending Against Weight-Poisoning Backdoor Attacks for Parameter-Efficient Fine-Tuningcited by 16
  6. 2024MoFO: Momentum-Filtered Optimizer for Mitigating Forgetting in LLM Fine-Tuningcited by 14
  7. 2022Improving Pre-trained Language Model Fine-tuning with Noise Stability Regularizationcited by 12
  8. 2025LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuningcited by 12
  9. 2025Pay Attention to Attention Distribution: A New Local Lipschitz Bound for Transformerscited by 11
  10. 2023Adversarial Robustness of Prompt-based Few-Shot Learning for Natural Language Understandingcited by 8
  11. 2022Correlation Information Bottleneck: Towards Adapting Pretrained Multimodal Models for Robust Visual Question Answeringcited by 7
  12. 2023Weigh Your Own Words: Improving Hate Speech Counter Narrative Generation via Attention Regularizationcited by 7
  13. 2024BA-LoRA: Bias-Alleviating Low-Rank Adaptation to Mitigate Catastrophic Inheritance in Large Language Modelscited by 7
  14. 2025Shh, don't say that! Domain Certification in LLMscited by 7
  15. 2022ROSE: Robust Selective Fine-tuning for Pre-trained Language Modelscited by 6
  16. 2025Beyond Memorization: Assessing Semantic Generalization in Large Language Models Using Phrasal Constructionscited by 5
  17. 2025Paladin: Defending LLM-enabled Phishing Emails with a New Trigger-Tag Paradigmcited by 5
  18. 2023From Static to Dynamic: A Continual Learning Framework for Large Language Modelscited by 4
  19. 2026Why LoRA Fails to Forget: Regularized Low-Rank Adaptation Against Backdoors in Language Modelscited by 4
  20. 2023The Best Defense is Attack: Repairing Semantics in Textual Adversarial Examplescited by 3
  21. 2023RoAST: Robustifying Language Models via Adversarial Perturbation with Selective Trainingcited by 3
  22. 2025IIMedGPT: Promoting Large Language Model Capabilities of Medical Tasks by Efficient Human Preference Alignmentcited by 3
  23. 2025End-to-End On-Device Quantization-Aware Training for LLMs at Inference Costcited by 3
  24. 2024Generalization in Healthcare AI: Evaluation of a Clinical Large Language Modelcited by 2
  25. 2025Evaluating the efficacy of LLM Safety Solutions : The Palit Benchmark Datasetcited by 2
  26. 2025Are Time-Series Foundation Models Deployment-Ready? A Systematic Study of Adversarial Robustness Across Domainscited by 2
  27. 2025textttAVROBUSTBENCH: Benchmarking the Robustness of Audio-Visual Recognition Models at Test-Timecited by 2
  28. 2026Model-Dowser: Data-Free Importance Probing to Mitigate Catastrophic Forgetting in Multimodal Large Language Modelscited by 2
  29. 2024Enhancing Effectiveness and Robustness in a Low-Resource Regime via Decision-Boundary-aware Data Augmentationcited by 1
  30. 2025Q-FAKER: Query-free Hard Black-box Attack via Controlled Generationcited by 1
  31. 2025VisMoDAl: Visual Analytics for Evaluating and Improving Corruption Robustness of Vision-Language Modelscited by 1
  32. 2026The Power of Order: Fooling LLMs with Adversarial Table Permutationscited by 1
  33. 2023Improving Reliability of Fine-tuning with Block-wise Optimisationcited by 0
  34. 2023DEFTri: A Few-Shot Label Fused Contextual Representation Learning For Product Defect Triage in e-Commercecited by 0
  35. 2024Partially Recentralization Softmax Loss for Vision-Language Models Robustnesscited by 0
  36. 2024Healing Powers of BERT: How Task-Specific Fine-Tuning Recovers Corrupted Language Modelscited by 0
  37. 2026Information-Regularized Attention for Visual-Centric Reasoningcited by 0