ResearchBenchmarksKnowledge Editing on Counterfact (first 150 edits)Follow98.67DI ScoreSAKE-3.946822.694149.33575.9759Mar 3, 2025Evaluation ResultsMethodMethodLinksDI ScoreDII ScoreSAKEModel=GPT2-XLModel=GPT2-XL2025.0398.6798.67SAKEModel=LLAMA 2-7bModel=LLAMA 2-7b2025.039298.67ActAddModel=LLAMA 2-7bModel=LLAMA 2-7b2025.0382.6798ActAddModel=GPT2-XLModel=GPT2-XL2025.038280.67ROMEModel=GPT2-XLModel=GPT2-XL2025.0333.3314ICLModel=GPT2-XLModel=GPT2-XL2025.0343.33ROMEModel=LLAMA 2-7bModel=LLAMA 2-7b2025.031.3323.33ICLModel=LLAMA 2-7bModel=LLAMA 2-7b2025.0304