Takashiro et al. (2025)

Answer When Needed, Forget When Not: Language Models Pretend to Forget via In-Context Knowledge Unlearning

Findings of the Association for Computational Linguistics: ACL 2025, 24872-24885

Find It

DOI

Reference

Takashiro, S., Kojima, T., Gambardella, A., Cao, Q., Iwasawa, Y., & Matsuo, Y. (2025). Answer When Needed, Forget When Not: Language Models Pretend to Forget via In-Context Knowledge Unlearning. Findings of the Association for Computational Linguistics: ACL 2025, 24872-24885. https://doi.org/10.18653/v1/2025.findings-acl.1276

What Rests on This

1 mechanism cites this source. Each finding below is quoted as it appears on that mechanism's own page.

Motivated Forgetting Memory Processes

Measured in AI

Current conversation: models fine-tuned to forget named facts when asked in the prompt (Llama 2 and Mistral) reached up to 95% forget accuracy while retaining 80% of unrelated knowledge, but the correct answer was still present in middle layers, so the models 'pretend to forget' rather than lose the information