Tuning Language Models by Proxy
Tuning Language Models by Proxy
Submitted on 16 Jan 2024 Tuning Language Models by Proxy
intuitive yet effective
- Proxy-tuning: A method to customize large language models without accessing their internal weights, by using smaller tuned models as experts and anti-experts.
- Premise: same dictionary
tunes a smaller LM, then applies the difference between the predictions of the small tuned and untuned LMs to shift the original predictions of the base model in the direction of tuning, while retaining the benefits of larger-scale pretraining.
