Abstract
Lifelong editing of LLMs requires storing thousands of edits after acquisition. A widely used family of approaches attaches one LoRA adapter per edit, which preserves behavior but grows linearly in storage. To address this challenge, we propose LadderEdit, a method that compresses each LoRA adapter after it is acquired. Each edit is first stored at low rank as a cheap sketch. We then check whether this sketch still satisfies the rewrite, generalization, and locality contract on probe prompts. Edits that pass keep the sketch; those that fail are promoted to a higher rank along a ladder until the contract is met. Because every edit retains some representation, coverage is maintained, and only hard edits consume more rank. Across ZsRE, CounterFact, and WikiBigEdit benchmarks on LLaMA-3-8B, Mistral-7B, and Qwen2.5-7B, LadderEdit tracks exact LoRA storage at 5.2x less memory and remains effective at 50,000 sequential edits.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Yu, X., Qiu, P., Yang, J., Dong, X., Ma, W., An, Z., Zhao, X., & Liu, X. (2026). LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs. https://omanscience.com/en/articles/ladderedit-edit-level-residual-compression-for-memory-efficient-lifelong-editing-of-llms
MLA 9
Yu, Xiaobing, et al. "LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs." https://omanscience.com/en/articles/ladderedit-edit-level-residual-compression-for-memory-efficient-lifelong-editing-of-llms.
Chicago (author–date)
Yu, Xiaobing, Peijie Qiu, Jin Yang, Xuanzhao Dong, Weiwei Ma, Zhaoqi An, Xiaoqi Zhao, and Xiaofeng Liu. 2026. "LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs." https://omanscience.com/en/articles/ladderedit-edit-level-residual-compression-for-memory-efficient-lifelong-editing-of-llms.
Harvard
Yu, X., Qiu, P., Yang, J., Dong, X., Ma, W., An, Z., Zhao, X. and Liu, X. (2026) 'LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs', Available at: https://omanscience.com/en/articles/ladderedit-edit-level-residual-compression-for-memory-efficient-lifelong-editing-of-llms.
Vancouver
Yu X, Qiu P, Yang J, Dong X, Ma W, An Z, et al. LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs. https://omanscience.com/en/articles/ladderedit-edit-level-residual-compression-for-memory-efficient-lifelong-editing-of-llms
IEEE
X. Yu, P. Qiu, J. Yang, X. Dong, W. Ma, Z. An, X. Zhao, and X. Liu, "LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs," https://omanscience.com/en/articles/ladderedit-edit-level-residual-compression-for-memory-efficient-lifelong-editing-of-llms.