Can LLMs forget to do something if intervening context is distracting? For instance, if I say "start the ninth sentence of your response with the word 'Eggplant'" how easy is it to make an LLM forget to do this? We should disentangle this from counting somehow.
If you are inspired by this idea, you can reach out to the authors for collaboration or cite it:
@misc{holtzman-forgetting-in-llm-2026,
author = {Holtzman, Ari},
title = {'Forgetting' in LLM Inference},
year = {2026},
url = {https://hypogenic.ai/ideahub/idea/wnvUvwO5kfmNxNNaxo3L}
}Please sign in to comment on this idea.
No comments yet. Be the first to share your thoughts!