# Suppressing Pink Elephants with Direct Principle Feedback

arXiv: https://arxiv.org/abs/2402.07896 · HTML: https://mlsanity.com/paper/2402.07896/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- Placed in Process reward models for reasoning, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 46 of its 64 references that are in the map.
- First posted on arXiv 12 Feb 2024.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Process reward models for reasoning](https://mlsanity.com/alignment/process-reward-models-for-reasoning/)

## Builds on (most cited first, 30 of 46)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Multitask Prompted Training Enables Zero-Shot Task Generalization](https://mlsanity.com/paper/2110.08207/) (2021)
- [Mass-Editing Memory in a Transformer](https://mlsanity.com/paper/2210.07229/) (2022)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Large Language Models Can Be Easily Distracted by Irrelevant Context](https://mlsanity.com/paper/2302.00093/) (2023)
- [WizardCoder: Empowering Code Large Language Models with Evol-Instruct](https://mlsanity.com/paper/2306.08568/) (2023)
- [WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct](https://mlsanity.com/paper/2308.09583/) (2023)
- [RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback](https://mlsanity.com/paper/2309.00267/) (2023)
- [Improving alignment of dialogue agents via targeted human judgements](https://mlsanity.com/paper/2209.14375/) (2022)
- [Neural Text Generation with Unlikelihood Training](https://mlsanity.com/paper/1908.04319/) (2019)
- [Reinforced Self-Training (ReST) for Language Modeling](https://mlsanity.com/paper/2308.08998/) (2023)
- [Who's Harry Potter? Approximate Unlearning in LLMs](https://mlsanity.com/paper/2310.02238/) (2023)
- [Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models](https://mlsanity.com/paper/2312.06585/) (2023)

## Built on by (most cited first, 15 of 15)

- [Social Choice Should Guide AI Alignment in Dealing with Diverse Human Feedback](https://mlsanity.com/paper/2404.10271/) (2024)
- [PERSONA: A Reproducible Testbed for Pluralistic Alignment](https://mlsanity.com/paper/2407.17387/) (2024)
- [Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2504.12501/) (2025)
- [Controllable Safety Alignment: Inference-Time Adaptation to Diverse Safety Requirements](https://mlsanity.com/paper/2410.08968/) (2024)
- [RLVF: Learning from Verbal Feedback without Overgeneralization](https://mlsanity.com/paper/2402.10893/) (2024)
- [CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialogues](https://mlsanity.com/paper/2404.03820/) (2024)
- [Configurable Safety Tuning of Language Models with Synthetic Preference Data](https://mlsanity.com/paper/2404.00495/) (2024)
- [Self-Directed Synthetic Dialogues and Revisions Technical Report](https://mlsanity.com/paper/2407.18421/) (2024)
- [Negation: A Pink Elephant in the Large Language Models' Room?](https://mlsanity.com/paper/2503.22395/) (2025)
- [Scalable Valuation of Human Feedback through Provably Robust Model Alignment](https://mlsanity.com/paper/2505.17859/) (2025)
- [Feedback That Backfires: Why Small Language Model Agents Repeat the Call They Just Watched Fail](https://mlsanity.com/paper/2608.23651/) (2026)
- [MISR: Measuring Instrumental Self-Reasoning in Frontier Models](https://mlsanity.com/paper/2412.03904/) (2024)
- [Antislop: A Comprehensive Framework for Identifying and Eliminating Repetitive Patterns in Language Models](https://mlsanity.com/paper/2510.15061/) (2025)
- [EvalSafetyGap: A Hybrid Survey and Conceptual Framework for LLM Evaluation-Safety Failures](https://mlsanity.com/paper/2606.30219/) (2026)
- [Visible but Not Avoided: State-Dependent Re-Proposal of Rejected Candidates in Catalog-Based LLM Optimization Loops](https://mlsanity.com/paper/2610.15738/) (2026)

## Abstract (arXiv)

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.