Research map / Large language models / Retrieval and knowledge
TangleScore: Tangle-Guided Purge and Imprint for Unstructured Knowledge Editing
Hao-Xiang Xu, Ziqi Peng, Jun-Yu Ma, Yuhao Sun, Zhen-Hua Ling, Jia-Chen Gu
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 |
|---|---|---|---|
| use ratio form of TangleScore instead of numerator only (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 32.79 -> 50.91 | Improved |
| use ratio form of TangleScore instead of denominator only (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 36.55 -> 50.91 | Improved |
| use dynamic purge rate instead of fixed 0.001 (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 48.21 -> 50.91 | Improved |
| use dynamic purge rate instead of fixed 0.0001 (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 43.85 -> 50.91 | Improved |
| add Lpurge loss (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 47.15 -> 50.91 | Improved |
| add Lconsistency loss (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 26.61 -> 50.91 | Improved |
| edit layer 7 instead of layer 4 (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 48.27 -> 50.91 | Improved |
| edit layer 7 instead of layer 10 (ablation) | PIPE · UNKEBench LLaMA3-8B-Instruct | FactScore: 48.11 -> 50.91 | Improved |
| use PIPE instead of AnyEdit | AnyEdit · UNKEBench LLaMA3-8B-Instruct | Rouge-2 (paraphrased): 72.59 -> 78.03 | Improved |
| use PIPE instead of UNKE | UNKE · KEBench | Para-Acc: 85.34 -> 88.23 | Improved |
| use PIPE instead of AnyEdit | AnyEdit · UNKEBench Qwen2.5-7B-Instruct | FactScore: 40.60 -> 42.47 | 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.
- AlphaEdit — AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025) · better on 2 of 2 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
- Uncovering Overfitting in Large Language Model Editing (ICLR 2025 · similarity 0.90)
- UniEdit: A Unified Knowledge Editing Benchmark for Large Language Models (NeurIPS 2025 · similarity 0.89)
- AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025 · similarity 0.89)
- Unveiling the Pitfalls of Knowledge Editing for Large Language Models (ICLR 2024 · similarity 0.89)
- Everything is Editable: Extend Knowledge Editing to Unstructured Data in Large Language Models (ICLR 2025 · similarity 0.88)
- Should We Really Edit Language Models? On the Evaluation of Edited Language Models (NeurIPS 2024 · similarity 0.88)
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.
