Research map / Large language models / Retrieval and knowledge
Multi-granularity Temporal Knowledge Editing over Large Language Models
Simiao Zhao, Ning Pang, Zhen Tan, Yanli Hu, Weidong Xiao, Xiang Zhao
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 time-anchored mechanism (timestamp anchors) (ablation) | SPIKE · MTKE-SG | ACC current: 50.11 -> 84.37 | Improved |
| add time-anchored mechanism (timestamp anchors) (ablation) | SPIKE · MTKE-MG | ACC current: 42.82 -> 75.69 | Improved |
| add time-anchored mechanism (ablation) | SPIKE · MTKE-SG | HTC: 0.6884 -> 0.7926 | Improved |
| add time-anchored mechanism (ablation) | SPIKE · MTKE-MG | HTC: 0.6640 -> 0.6904 | Improved |
| add time-anchored mechanism (ablation) | SPIKE · MTKE-SG | ACC history: 82.66 -> 85.01 | Improved |
| use SPIKE instead of best METO-based baseline | SPIKE (GPT-J) · MTKE-SG | ACC history: 26.55 -> 85.01 | Improved |
| use SPIKE instead of best METO-based baseline | SPIKE (GPT-J) · MTKE-MG | ACC history: 29.83 -> 80.66 | Improved |
| use SPIKE instead of best METO-based baseline | SPIKE (GPT-J) · MTKE-SG | HTC: 0.5734 -> 0.7926 | Improved |
| use SPIKE instead of ROME | SPIKE (GPT-J) · MTKE-SG | ACC current: 91.22 -> 84.37 | Hurt |
| use SPIKE instead of ROME | SPIKE (GPT-J) · MTKE-MG | ACC current: 92.27 -> 75.69 | Hurt |
| use SPIKE instead of ROME | SPIKE (GPT-J) · MTKE-SG | Soft Recall: 0.4820 -> 0.3552 | Hurt |
| filter AToKe samples where model already knows target or lacks history | SPIKE · AToKe-SE | ACC history: 20.25 -> 29.86 | Improved |
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.
- MEND† — MEND: Meta Demonstration Distillation for Efficient and Effective In-Context Learning (ICLR 2024) · better on 7 of 8 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
- History Matters: Temporal Knowledge Editing in Large Language Model (AAAI 2024 · similarity 0.93)
- WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models (NeurIPS 2024 · similarity 0.88)
- Large Language Models-guided Dynamic Adaptation for Temporal Knowledge Graph Reasoning (NeurIPS 2024 · similarity 0.87)
- ChroKnowledge: Unveiling Chronological Knowledge of Language Models in Multiple Domains (ICLR 2025 · similarity 0.87)
- Uncovering Overfitting in Large Language Model Editing (ICLR 2025 · similarity 0.86)
- AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025 · 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.
