Work

Detecting Edit Failures In Large Language Models: An Improved Specificity Benchmark

Year: 2023
Type: conference-paper
Abstract: Recent model editing techniques promise to mitigate the problem of memorizing false or outdated associations during large language model (LLM) training.However, we show that these techniques can introduce large unwanted side effects which are not detected by existing specificity benchmarks.We extend the existing COUNTERFACT benchmark to include a dynamic component and dub our benchmark COUNTERFACT+.Additionally, we extend the metrics used for measuring specificity by a principled KL divergence-based metric.We use this improved benchmark to evaluate recent model editing techniques and find that they suffer from low specificity.Our findings highlight the need for improved specificity benchmarks that identify and prevent unwanted side effects.
Language: English
Open Access status: hybrid
OpenAlex ID: W4385571289
FWCI: 4.851
Cites: 15
Cited by: 83
Related to: 10