mlsanityData to 7 Oct 2026

Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation

arXiv 2603.13683 ↗First posted 14 Mar 2026cs.CL cs.AI cs.CYcited by 5

Not on the map. Why?

Abstract (arXiv)

Although debiased large language models (LLMs) excel at handling known or low-bias prompts, they often fail on unfamiliar and high-bias prompts. We demonstrate via out-of-distribution (OOD) detection that these high-bias prompts cause a distribution shift, degrading static model performance. To enable real-time correction, we propose CAP-TTA, a test-time adaptation framework. CAP-TTA triggers context-aware LoRA updates only when a bias-risk score exceeds a set threshold. By utilizing an offline precomputed diagonal preconditioner, it ensures fast and stable optimization. Across multiple benchmarks and human evaluations, CAP-TTA effectively reduces toxicity/bias score with significantly lower latency than standard optimization methods (e.g., AdamW or SGD). Furthermore, it prevents catastrophic forgetting, and substantially improves narrative fluency over state-of-the-art baselines without compromising debiasing performance.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Builds on 41 of its 67 references that are in the map. First posted on arXiv 14 Mar 2026.

Builds on

41 of 67 references.

References · 41

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  4. 2016Overcoming catastrophic forgetting in neural networkscited by 4,718
  5. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,355
  6. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  7. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,472
  8. 2018Continual Lifelong Learning with Neural Networks: A Reviewcited by 1,313
  9. 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
  10. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  11. 2018Hierarchical Neural Story Generationcited by 1,182
  12. 2023Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!cited by 1,106
29 more
  1. 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 951
  2. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 911
  3. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 793
  4. 2015Optimizing Neural Networks with Kronecker-factored Approximate Curvaturecited by 776
  5. 2020CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Modelscited by 703
  6. 2018Mitigating Unwanted Biases with Adversarial Learningcited by 663
  7. 2018Gender Bias in Coreference Resolution: Evaluation and Debiasing Methodscited by 657
  8. 2022Out-of-Distribution Detection with Deep Nearest Neighborscited by 596
  9. 2019The Woman Worked as a Babysitter: On Biases in Language Generationcited by 540
  10. 2024Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Trainingcited by 478
  11. 2019Likelihood Ratios for Out-of-Distribution Detectioncited by 451
  12. 2020Null It Out: Guarding Protected Attributes by Iterative Nullspace Projectioncited by 442
  13. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 434
  14. 2017Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraintscited by 420
  15. 2023Towards Stable Test-Time Adaptation in Dynamic Wild Worldcited by 409
  16. 2021MEMO: Test Time Robustness via Adaptation and Augmentationcited by 376
  17. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 374
  18. 2021Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLPcited by 336
  19. 2020Evaluation of Text Generation: A Surveycited by 275
  20. 2019A Framework for Understanding Sources of Harm throughout the Machine Learning Life Cyclecited by 192
  21. 2023SteerLM: Attribute Conditioned SFT as an (User-Steerable) Alternative to RLHFcited by 107
  22. 2024A Framework for Human Evaluation of Large Language Models in Healthcare Derived from Literature Reviewcited by 65
  23. 2022The Authenticity Gap in Human Evaluationcited by 24
  24. 2025Prompt-based Adaptation in Large-scale Vision Models: A Surveycited by 21
  25. 2024Large Language Models Still Exhibit Bias in Long Textcited by 16
  26. 2023ATTA: Anomaly-aware Test-Time Adaptation for Out-of-Distribution Detection in Segmentationcited by 14
  27. 2024What is the Best Way for ChatGPT to Translate Poetry?cited by 14
  28. 2025Benchmarking the Detection of LLMs-Generated Modern Chinese Poetrycited by 10
  29. 2025Depth-Wise Activation Steering for Honest Language Modelscited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Not All Directions Matter: Towards Structured and Task-Aware Low-Rank Model Adaptationcited by 11
  2. 2026Can ChatGPT Really Understand Modern Chinese Poetry?cited by 4
  3. 2026Seeing the Poem: Image-Semantic Detection of AI-Generated Modern Chinese Poetry with MLLMscited by 1
  4. 2026BV-Blend: Uncertainty-Weighted Historical Baselines for Stable Critic-Free RL with Verifiable Rewardscited by 0
  5. 2026Poller: Are LLMs Suitable for Evaluating the Poetry Understanding Task?cited by 0