paper-with-me

홈 › Papers

MEMIT-Merge: Addressing MEMIT's Key-Value Conflicts in Same-Subject Batch Editing for LLMs

2025-02-11 · Zilu Dong, Xiangqing Shen, Rui Xia

As large language models continue to scale up, knowledge editing techniques that modify models' internal knowledge without full retraining have gained significant attention. MEMIT, a prominent batch editing algorithm, stands out for its capability to perform mass knowledge modifications. However, we uncover a critical limitation that MEMIT's editing efficacy significantly deteriorates when processing batches containing multiple edits sharing the same subject. Our analysis reveals that the root cause lies in MEMIT's key value modeling framework: When multiple facts with the same subject in a batch are modeled through MEMIT's key value mechanism, identical keys (derived from the shared subject) are forced to represent different values (corresponding to different knowledge), resulting in updates conflicts during editing. Addressing this issue, we propose MEMIT-Merge, an enhanced approach that merges value computation processes for facts sharing the same subject, effectively resolving the performance degradation in same-subject batch editing scenarios. Experimental results demonstrate that when MEMIT's edit success rate drops to around 50% at larger batch sizes, MEMIT-Merge maintains a success rate exceeding 90%, showcasing remarkable robustness to subject entity collisions.

📄 PDF Abstract BibTeX arXiv:2502.07322

Code (0)

등록된 구현이 없습니다.

Tasks

knowledge editing

Similar Papers 제목 키워드 기반

A Unified Framework for Model Editing

2024-03-21 · Akshat Gupta, Dev Sajnani, Gopala Anumanchipalli

ROME and MEMIT are largely believed to be two different model editing algorithms, with the major difference between them being the ability to perform batched edits. In this paper, we unify these two algorithms under a si…

MemorizationmodelModel Editing

Editing Common Sense in Transformers

2023-05-24 · Anshita Gupta, Debanjan Mondal, Akshay Krishna Sheshadri, Wenlong Zhao 외

Editing model parameters directly in Transformers makes updating open-source transformer-based models possible without re-training (Meng et al., 2023). However, these editing methods have only been evaluated on statement…

Common Sense ReasoningModel EditingSpecificity

Mass-Editing Memory in a Transformer

2022-10-13 · Kevin Meng, Arnab Sen Sharma, Alex Andonian, Yonatan Belinkov 외

Recent work has shown exciting promise in updating large language models with new memories, so as to replace obsolete information or add specialized knowledge. However, this line of work is predominantly limited to updat…

Language ModelingLanguage Modelling

One Mask to Rule Them All: On Hidden Facts after Editing and How to Find Them

2026-04-18 · Ali Holmov, Paul Youssef, Nandi Schoots, Christin Seifert arxiv

Knowledge editing methods such as ROME and MEMIT update factual associations in transformer models by modifying MLP weights. While evaluated mainly by output behavior, their internal mechanism remains underexplored. We i…

knowledge editing

Scalable Knowledge Editing for Mixture-of-Experts LLMs via Tensor-Structured Updates

2026-05-15 · Roman Maksimov, Vladimir Aletov, Dmitry Bylinkin, Daniil Medyakov 외 arxiv

Knowledge editing (KE) provides a lightweight alternative to repeated fine-tuning of LLMs. However, most existing KE methods target dense feed-forward layers, while modern LLMs increasingly adopt Mixture-of-Experts (MoE)…

knowledge editing