mlsanityData to 14 Oct 2026

Popular Knowledge Propagates More Errors in LLM Knowledge Updating

arXiv 2609.08067 ↗First posted 8 Sep 2026cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Updating a language model's knowledge through fine-tuning is essential for keeping its outputs current, yet can also induce factual forgetting and new hallucinations. Prior work shows that long-tail knowledge is harder to acquire and newly memorized long-tail facts are difficult to retain during later fine-tuning. We study a complementary question: among facts that a model has encoded correctly, which are most vulnerable to collateral corruption during other updates? To investigate this question under a realistic factual distribution, we construct a large-scale graph FACTPROP of verified Wikipedia facts by linking triples that share head or tail entities, thereby preserving connections among factual knowledge. We fine-tune models on factual statements and measure correct-to-incorrect facts after each update. Our results reveal a pattern distinct from prior findings on long-tail vulnerability during acquisition and retention: among facts that models already answer correctly, those associated with highly connected entities are more likely to be corrupted by neighboring updates, and updates to such facts propagate errors more broadly. Structural popularity therefore predicts both vulnerability and downstream damage. Inspired by this finding, we propose Popularity-based Anchoring (PopAnchor), a lightweight rehearsal strategy that preserves a small set of popular facts and reduces forgetting.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 31 of its 41 references that are in the map. First posted on arXiv 8 Sep 2026.

Builds on

31 of 41 references.

References · 31

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  2. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  3. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  4. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  5. 2017Gradient Episodic Memory for Continual Learningcited by 1,898
  6. 2022When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memoriescited by 1,128
  7. 2018Efficient Lifelong Learning with A-GEMcited by 1,058
  8. 2020Dark Experience for General Continual Learning: a Strong, Simple Baselinecited by 911
  9. 2023An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuningcited by 639
  10. 2019Gradient based sample selection for online continual learningcited by 571
  11. 2023Dissecting Recall of Factual Associations in Auto-Regressive Language Modelscited by 528
  12. 2022Large Language Models Struggle to Learn Long-Tail Knowledgecited by 458
19 more
  1. 2019Online Continual Learning with Maximally Interfered Retrievalcited by 379
  2. 2023Editing Large Language Models: Problems, Methods, and Opportunitiescited by 372
  3. 2023Orthogonal Subspace Learning for Language Model Continual Learningcited by 312
  4. 2023MQuAKE: Assessing Knowledge Editing in Language Models via Multi-Hop Questionscited by 300
  5. 2024Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?cited by 232
  6. 2020Modifying Memories in Transformer Modelscited by 224
  7. 2019Episodic Memory in Lifelong Language Learningcited by 188
  8. 2021Towards Continual Knowledge Learning of Language Modelscited by 174
  9. 2020Recall and Learn: Fine-tuning Deep Pretrained Language Models with Less Forgettingcited by 170
  10. 2024Mitigating Catastrophic Forgetting in Large Language Models with Self-Synthesized Rehearsalcited by 118
  11. 2024Revisiting Catastrophic Forgetting in Large Language Model Tuningcited by 99
  12. 2024Model Editing at Scale leads to Gradual and Catastrophic Forgettingcited by 93
  13. 2024Rehearsal-Free Modular and Compositional Continual Learning for Language Modelscited by 27
  14. 2025WikiBigEdit: Understanding the Limits of Lifelong Knowledge Editing in LLMscited by 26
  15. 2024Propagation and Pitfalls: Reasoning-based Assessment of Knowledge Editing through Counterfactual Taskscited by 24
  16. 2024Why Does New Knowledge Create Messy Ripple Effects in LLMs?cited by 18
  17. 2026Replaying pre-training data improves fine-tuningcited by 13
  18. 2024Continual Memorization of Factoids in Language Modelscited by 8
  19. 2024An Efficient Rehearsal Scheme for Catastrophic Forgetting Mitigation during Multi-stage Fine-tuningcited by 4

Built on by

Not cited yet.