ACL Findings2024
Propagation and Pitfalls: Reasoning-based Assessment of Knowledge Editing through Counterfactual Tasks
Wenyue Hua, Jiang Guo, Mingwen Dong, Henghui Zhu, Patrick Ng, Zhiguo Wang
Abstract
Current knowledge editing approaches struggle to effectively propagate updates to interconnected facts.In this work, we delve into the barriers that hinder the appropriate propagation of updated knowledge within these models for accurate reasoning.To support our analysis, we introduce a novel reasoningbased benchmark, ReCoE (Reasoning-based Counterfactual Editing dataset), which covers six common reasoning schemes in the real world.We conduct an extensive analysis of existing knowledge editing techniques, including input-augmentation, finetuning, and locate-andedit methods.We found that all model editing methods exhibit notably low performance on this dataset, especially within certain reasoning schemes.Our analysis of the chain-ofthought responses from edited models indicate that, while the models effectively update individual facts, they struggle to recall these facts in reasoning tasks.Moreover, locate-and-edit methods severely deteriorate the models' language modeling capabilities, leading to poor perplexity and logical coherence in their outputs.Pidgeon Island belongs to Antarctica Asia Q: Which continent does Pidgeon Island belong to?Q: Is Pidgeon Island in the same continent as China?Let's think step by step: Asia is the same continent as China.Asia is the largest continent in the world [Irrelevant].Asia includes Australia [Incorrect].So the answer is yes [Incoherent]. After Edit (MEMIT) After Edit (expect)Pidgeon Island belongs to Asia.China also belongs to Asia.So the answer is yes. fact update Direct ReasoningBefore Edit Pidgeon Island is a rocky island located in Antarctica.China is a country in Asia.So the answer is no. Before Edit Antarctica After Edit Asia This workAntarctica Asia Antarctica Asia
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
13 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Keys to Robust Edits: From Theoretical Insights to Practical Advances cites Memory-Based Model Editing at Scale
- Keys to Robust Edits: From Theoretical Insights to Practical Advances cites Unveiling the Pitfalls of Knowledge Editing for Large Language Models
- QLoRA: Efficient Finetuning of Quantized LLMs cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- Can Editing LLMs Inject Harm? cites Unveiling the Pitfalls of Knowledge Editing for Large Language Models
- Unveiling the Pitfalls of Knowledge Editing for Large Language Models cites Memory-Based Model Editing at Scale
- Unveiling the Pitfalls of Knowledge Editing for Large Language Models cites Fast Model Editing at Scale
- Dynamic Retriever for In-Context Knowledge Editing via Policy Optimization cites Unveiling the Pitfalls of Knowledge Editing for Large Language Models
- Improving Factuality and Reasoning in Language Models through Multiagent Debate cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Cited by top-tier papers13
- Unveiling the Pitfalls of Knowledge Editing for Large Language ModelsZhoubo Li, Ningyu Zhang, Yunzhi Yao, Mengru Wang et al.ICLR 2024 · 47 citations
- Can Editing LLMs Inject Harm?Canyu Chen, Baixiang Huang, Zekun Li, Zhaorun Chen et al.AAAI 2026 · 26 citations
- Keys to Robust Edits: From Theoretical Insights to Practical AdvancesJianhao Yan, Futing Wang, Yun Luo, Yafu Li et al.ACL 2025 · 3 citations
- KnowTuning: Knowledge-aware Fine-tuning for Large Language ModelsYougang Lyu, Lingyong Yan, Shuaiqiang Wang, Haibo Shi et al.EMNLP 2024 · 3 citations
- Dynamic Retriever for In-Context Knowledge Editing via Policy OptimizationMahmud Wasif Nafee, Maiqi Jiang, Haipeng Chen, Yanfu ZhangEMNLP 2025 · 3 citations
Builds on16
- Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma et al.NeurIPS 2022 · 22,562 citations
- QLoRA: Efficient Finetuning of Quantized LLMsTim Dettmers, Artidoro Pagnoni, Ari Holtzman, Luke ZettlemoyerNeurIPS 2023 · 5,863 citations
- Improving Factuality and Reasoning in Language Models through Multiagent DebateYilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum et al.ICML 2024 · 1,562 citations
- Fast Model Editing at ScaleEric Mitchell, Charles Lin, Antoine Bosselut, Chelsea Finn et al.ICLR 2022 · 527 citations
- Memory-Based Model Editing at ScaleEric Mitchell, Charles Lin, Antoine Bosselut, Christopher D. Manning et al.ICML 2022 · 520 citations
Related papers
- Outdated Issue Aware Decoding for Factual Knowledge EditingZengkui Sun, Yijin Liu, Jiaan Wang, Fandong Meng et al.ACL Findings 2024
- RelEdit: Evaluating Conceptual Knowledge Editing in Language Models via Relational ReasoningYifan Niu, Miao Peng, Nuo Chen, Yatao Bian et al.ACL Findings 2025
- Decoupling Reasoning and Knowledge Injection for In-Context Knowledge EditingChangyue Wang, Weihang Su, Qingyao Ai, Yujia Zhou et al.ACL Findings 2025 · 2 citations
- Uncovering Overfitting in Large Language Model EditingMengqi Zhang, Xiaotian Ye, Qiang Liu, Shu Wu et al.ICLR 2025
- CaKE: Circuit-aware Editing Enables Generalizable Knowledge LearnersYunzhi Yao, Jizhan Fang, Jia-Chen Gu, Ningyu Zhang et al.EMNLP 2025 · 1 citation
