Research map / Large language models / Retrieval and knowledge
Editing Is a Bargaining Game: Balanced Knowledge Editing in Large Language Models
Chenghao Xu, Jiexi Yan, Muli Yang, Fen Fang, Huilin Chen, Cheng Deng
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 |
|---|---|---|---|
| replace naive gradient descent with Nash bargaining dynamic coefficients | UnKE (GPT-J) · WikiDataCounterfact | Forgetfulness Accuracy: 40.1 -> 63.5 | Improved |
| replace naive gradient descent with Nash bargaining | UnKE (Qwen2.5-7B-Instruct) · WikiDataCounterfact | Forgetfulness Accuracy: 48.6 -> 66.8 | Improved |
| replace naive gradient descent with Nash bargaining | UnKE (Llama3-8B-Instruct) · WikiDataCounterfact | Forgetfulness Accuracy: 44.1 -> 64.7 | Improved |
| replace naive gradient descent with Nash bargaining | UnKE (Qwen2.5-7B-Instruct) · WikiDataRecent | Forgetfulness Accuracy: 59.4 -> 74.2 | Improved |
| replace naive gradient descent with Nash bargaining | UnKE (GPT-J) · WikiDataRecent | Subject Aliasing Accuracy: 87.6 -> 92.1 | Improved |
| replace naive gradient descent with Nash bargaining | UnKE (Qwen2.5-7B-Instruct) · WikiDataCounterfact | Subject Aliasing Accuracy: 76.8 -> 85.3 | Improved |
| replace naive gradient descent with Nash bargaining | UnKE · WikiDataCounterfact (Llama3-8B) | Edit Success: 99.7 -> 99.7 | No clear effect |
| use Nash bargaining NaKE instead of AlphaEdit | AlphaEdit (GPT-J) · WikiDataCounterfact | Forgetfulness Accuracy: 55.7 -> 63.5 | Improved |
| use dynamic Nash coefficients instead of fixed lambda_p (1 to 10) with lambda_e=1 (ablation) | UnKE · Coefficient sweep (Figure 4) | preservation vs editing | Mixed |
| add dynamic coefficient updating (ablation) | UnKE · Editing time and memory | editing time | Hurt |
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 4 of 4 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 0.88)
- Mitigating Heterogeneous Token Overfitting in LLM Knowledge Editing (ICML 2025 · similarity 0.88)
- Golden Layers and Where to Find Them: Improved Knowledge Editing for Large Language Models Via Layer Gradient Analysis (NeurIPS 2026 · similarity 0.87)
- Unveiling the Pitfalls of Knowledge Editing for Large Language Models (ICLR 2024 · similarity 0.87)
- UniEdit: A Unified Knowledge Editing Benchmark for Large Language Models (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.
