# MidPO: Dual Preference Optimization for Safety and Helpfulness in Large Language Models via a Mixture of Experts Framework

arXiv: https://arxiv.org/abs/2506.02460 · HTML: https://mlsanity.com/paper/2506.02460/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 18 of its 38 references that are in the map.
- First posted on arXiv 3 Jun 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Safety alignment and jailbreak attacks](https://mlsanity.com/alignment/safety-alignment-and-jailbreak-attacks/)
- Mixture-of-experts / [MoE large language models](https://mlsanity.com/moe/moe-large-language-models/)

## Builds on (most cited first, 18 of 18)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [LaMDA: Language Models for Dialog Applications](https://mlsanity.com/paper/2201.08239/) (2022)
- [DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence](https://mlsanity.com/paper/2401.14196/) (2024)
- [DeepSeek-V3 Technical Report](https://mlsanity.com/paper/2412.19437/) (2024)
- [RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment](https://mlsanity.com/paper/2304.06767/) (2023)
- [Personalized Soups: Personalized Large Language Model Alignment via Post-hoc Parameter Merging](https://mlsanity.com/paper/2310.11564/) (2023)
- [PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference](https://mlsanity.com/paper/2406.15513/) (2024)
- [OR-Bench: An Over-Refusal Benchmark for Large Language Models](https://mlsanity.com/paper/2405.20947/) (2024)
- [Disentangling Length from Quality in Direct Preference Optimization](https://mlsanity.com/paper/2403.19159/) (2024)
- [Rule Based Rewards for Language Model Safety](https://mlsanity.com/paper/2411.01111/) (2024)
- [Merging Multi-Task Models via Weight-Ensembling Mixture of Experts](https://mlsanity.com/paper/2402.00433/) (2024)
- [Enhancing LLM Safety via Constrained Direct Preference Optimization](https://mlsanity.com/paper/2403.02475/) (2024)
- [Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models](https://mlsanity.com/paper/2408.15313/) (2024)
- [MoGU: A Framework for Enhancing Safety of Open-Sourced LLMs While Preserving Their Usability](https://mlsanity.com/paper/2405.14488/) (2024)

## Built on by (most cited first, 3 of 3)

- [RASA: Routing-Aware Safety Alignment for Mixture-of-Experts Models](https://mlsanity.com/paper/2602.04448/) (2026)
- [Provably Convergent Primal-Dual DPO for Constrained LLM Alignment](https://mlsanity.com/paper/2510.05703/) (2025)
- [Where Paths Split: Localized, Calibrated Control of Moral Reasoning in Large Language Models](https://mlsanity.com/paper/2605.03609/) (2026)

## Abstract (arXiv)

As large language models (LLMs) are increasingly applied across various domains, enhancing safety while maintaining the helpfulness of LLMs has become a critical challenge. Recent studies solve this problem through safety-constrained online preference optimization or safety-constrained offline preference optimization. However, the safety-constrained online methods often suffer from excessive safety, which might reduce helpfulness, while the safety-constrained offline methods perform poorly in adaptively balancing safety and helpfulness. To address these limitations, we propose MidPO, a textbfunderlineMixture of Experts (MoE) framework for safety-helpfulness textbfunderlinedual textbfunderlinePreference textbfunderlineOptimization. Firstly, MidPO devises single-preference enhanced direct preference optimization approach to transform the base model into two independent experts, termed safety and helpfulness experts, and fine-tunes the two independent experts for optimal safety or helpfulness performance. Secondly, to achieve an effective balance between safety and helpfulness, MidPO incorporates the two experts into the MoE framework and designs a dynamic routing mechanism to allocate contributions from each expert adaptively. We conduct quantitative and qualitative experiments on three popular datasets to demonstrate the proposed MidPO significantly outperforms state-of-the-art approaches in both safety and helpfulness. The code and models will be released.