Prompt Injection: Parameterization of Fixed Inputs

By Eunbi Choi, Yongrae Jo, Joel Jang, Minjoon Seo. 2022

42 citations recorded by the source corpus.

Recent works have shown that attaching prompts to the input is effective at conditioning Language Models (LM) to perform specific tasks. However, prompts are always included in the input text during inference, thus incurring substantial computational and memory overhead. Also, there is currently no straightforward method of utilizing prompts that are longer than the maximum input length of the LMs without incurring additional costs during inference. We propose Prompt Injection (PI), a novel formulation of injecting the prompt into the parameters of an LM to be an efficient alternative to attaching fixed prompts to the input. We show that in scenarios with long fixed prompts, PI can be up to 280 times more efficient in terms of total FLOPs than previous approaches. We further explore methodologies for PI and show promising results in persona-dependent conversation, semantic parsing, and zero-shot learning with task instructions. Through these explorations, we show that PI can be a promising direction for conditioning language models, especially in scenarios with long and fixed prompts.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Daniel Klein
2. Anca Dragan
3. Noah D. Goodman
4. Dorsa Sadigh
5. Mohit Bansal
6. Ruiqi Zhong
7. Percy Liang
8. Jesse Mu

In Safety, Robustness & Interpretability

1. Dylan Hadfield-Menell
2. Anca Dragan
3. Stephen Casper
4. Max Nadeau
5. Noah D. Goodman
6. Peter Hase
7. David Krueger
8. Jacob Pfau

Citations in the corpus

• Learning by Distilling Context (2022, methodology): “Beyond the above two variants, Choi et al. (2022) explores a recursive variant, which is similar to sequential distillation, except that the student model becomes the new teacher in the next iteration.”
• Learning to Compress Prompts with Gist Tokens (2023, methodology): “2 This is precisely the approach taken by recent work (Askell et al., 2021; Choi et al., 2022; Snell et al., 2022), including Wingate et al.”
• Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback (2023, background): “…that explicitly model multiple evaluators [Gordon et al., 2021, Davani et al., 2022, Daniels-Koch and Freedman, 2022, Gordon et al., 2022, Barnett et al., 2023], that tune models to individuals [Kumar et al., 2021], or that use more sophisticated aggregation strategies [Noothigattu et al., 2018].”