Research map / Large language models / Retrieval and knowledge
MobiEdit: Resource-efficient Knowledge Editing for Personalized On-device LLMs
Zhenyan Lu, Daliang Xu, Dongqi Cai, Zexi Li, Wei Liu, Jian Luan, Fangming Liu, Shangguang Wang, Mengwei Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add early stopping controller to zeroth-order editing (ablation) | MobiEdit zo · ZsRE | editing time: over 40% less time | Improved |
| add early stopping controller to zeroth-order editing (ablation) | MobiEdit zo · ZsRE | edit success | No clear effect |
| add prefix cache on top of early stopping (ablation) | MobiEdit · ZsRE | editing time: another 20-30% faster | Improved |
| combine early stopping and prefix cache (ablation) | basic zeroth-order editing · ZsRE | editing time: over 4000 seconds -> nearly one-third | Improved |
| use quantized forward-only editing on NPU instead of ROME on CPU | ROME · ZsRE (K60 Pro) | editing time (s): 4543.78 -> 1902.88 | Improved |
| use MobiEdit instead of ROME/MEMIT/AlphaEdit | ROME · ZsRE | memory (GB): 46.14 -> 6.20 | Improved |
| use MobiEdit instead of WISE | WISE · ZsRE (K60 Pro) | energy per edit (J): 0.63 -> 0.023 | Improved |
| use zeroth-order estimates instead of backpropagation | BP-based editing · ZsRE and CounterFact | number of steps: 20x more steps | Hurt |
| use MobiEdit instead of MEMIT and AlphaEdit | MEMIT/AlphaEdit · ZsRE and CounterFact | edit success: 13.9 reduction in edit success | Hurt |
| reuse stale prefix activations via prefix cache (ablation) | MobiEdit · ZsRE | cosine similarity of QKV representations: remains as high as 0.9 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- WISE — WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models (NeurIPS 2024) · better on 12 of 12 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs (NeurIPS 2025 · similarity 0.89)
- MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases (ICML 2024 · similarity 0.88)
- AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025 · similarity 0.86)
- UniEdit: A Unified Knowledge Editing Benchmark for Large Language Models (NeurIPS 2025 · similarity 0.86)
- Precise Localization of Memories: A Fine-grained Neuron-level Knowledge Editing Technique for LLMs (ICLR 2025 · similarity 0.86)
- WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
