Papers
arxiv:2611.15479

DETER: Suppression During Pretraining for Tamper Resistance

Published on Nov 11
Authors:
,

Abstract

Suppressing unwanted knowledge during pretraining, by interleaving gradient ascent with standard pretraining on filtered data, makes that knowledge more resistant to later fine-tuning than filtering alone or post-hoc unlearning.

Open-weight language models can acquire knowledge that could enable biological, chemical, or cyber misuse. Any attempt to limit such knowledge must remain effective after the model weights are released and can be fine-tuned. Unlearning applied after pretraining can be rapidly undone, whereas filtering the pretraining corpus offers greater resistance. We ask whether actively suppressing knowledge during pretraining can further increase the fine-tuning required to acquire it. Even simple gradient ascent interleaved with standard pretraining on the filtered corpus makes medical knowledge harder to acquire through adversarial fine-tuning than filtering alone or post-hoc unlearning. At 1B parameters, reaching the held-out PubMed loss of an unfiltered model requires over 20times as many PubMed fine-tuning tokens as after filtering alone. Retained knowledge is evaluated using both held-out data and downstream tasks. Across model sizes, varying suppression strength reveals an increasingly favorable trade-off between fine-tuning difficulty and held-out loss on retained pretraining data. These results point to pretraining as an effective stage for making unwanted capabilities harder to acquire through subsequent fine-tuning. Code is available at https://github.com/MEGlr/DETER{github.com/MEGlr/DETER}.

Community

Sign up or log in to comment

Get this paper in your agent:

hf papers read 2611.15479
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2611.15479 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.15479 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2611.15479 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.