Suppressing Pink Elephants with Direct Principle Feedback

TL;DR

This paper introduces Direct Principle Feedback, a new method for controlling language models at inference time to avoid discussing specific entities, demonstrated by outperforming existing models on the Pink Elephant problem.

Contribution

It proposes a novel simplification of Constitutional AI called Direct Principle Feedback, enabling effective inference-time control without response ranking.

Findings

DPF fine-tuning outperforms Llama-2-13B-Chat and baseline models.

The 13B LLaMA 2 model matches GPT-4 on Pink Elephant tests.

Method effectively controls LLM behavior at inference time.

Abstract

Existing methods for controlling language models, such as RLHF and Constitutional AI, involve determining which LLM behaviors are desirable and training them into a language model. However, in many cases, it is desirable for LLMs to be controllable at inference time, so that they can be used in multiple contexts with diverse needs. We illustrate this with the Pink Elephant Problem: instructing an LLM to avoid discussing a certain entity (a ``Pink Elephant''), and instead discuss a preferred entity (``Grey Elephant''). We apply a novel simplification of Constitutional AI, Direct Principle Feedback, which skips the ranking of responses and uses DPO directly on critiques and revisions. Our results show that after DPF fine-tuning on our synthetic Pink Elephants dataset, our 13B fine-tuned LLaMA 2 model significantly outperforms Llama-2-13B-Chat and a prompted baseline, and performs as well as GPT-4 in on our curated test set assessing the Pink Elephant Problem.