Research map / Large language models / Retrieval and knowledge
Memory-Based Model Editing at Scale
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace parameter-updating editors with memory-based SERAC | Model editing · QA | ES: 0.944 -> 0.986 | Helped |
| Replace parameter-updating editors with memory-based SERAC | Model editing · QA-hard | ES: 0.515 -> 0.913 | Helped |
| Replace parameter-updating editors with memory-based SERAC | Model editing · FC | ES: 0.627 -> 0.877 | Helped |
| Use SERAC instead of the strongest baseline | Model editing · FC | DD: 0.002 -> 0.051 | Hurt |
| Replace parameter-updating editors with memory-based SERAC | Model editing · ConvSent | ES: 0.506 -> 0.991 | Helped |
| Use cross-attention rather than embedding-based scope classification (ablation) | SERAC with distilBERT scope-classifier backbone · QA-hard | ES: 0.921 -> 0.983 | Helped |
| Use cross-attention rather than embedding-based scope classification (ablation) | SERAC with distilBERT scope-classifier backbone · FC | ES: 0.792 -> 0.831 | Helped |
| Use cross-attention rather than embedding-based scope classification (ablation) | SERAC with BERT-base scope-classifier backbone · QA-hard | ES: 0.945 -> 0.983 | Helped |
| Use cross-attention rather than embedding-based scope classification (ablation) | SERAC with BERT-base scope-classifier backbone · FC | ES: 0.792 -> 0.855 | Helped |
| Increase scope-classifier backbone from distilBERT to BERT-base with embedding classification (ablation) | SERAC · QA-hard | ES: 0.921 -> 0.945 | Helped |
| Increase scope-classifier backbone from distilBERT to BERT-base with embedding classification (ablation) | SERAC · FC | ES: 0.792 -> 0.792 | No clear effect |
| Increase scope-classifier backbone from distilBERT to BERT-base with cross-attention classification (ablation) | SERAC · QA-hard | ES: 0.983 -> 0.983 | No clear effect |
Closest papers by meaning
- WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models (NeurIPS 2024 · similarity 0.89)
- Should We Really Edit Language Models? On the Evaluation of Edited Language Models (NeurIPS 2024 · similarity 0.88)
- Massive Editing for Large Language Models via Meta Learning (ICLR 2024 · similarity 0.87)
- Editing Memories Through Few Targeted Neurons (AAAI 2025 · similarity 0.87)
- MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs (NeurIPS 2025 · similarity 0.87)
- AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025 · similarity 0.87)
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.