HIPO: Instruction Hierarchy via Constrained Reinforcement Learning

TL;DR

HIPO introduces a constrained reinforcement learning framework that enforces strict adherence to system prompts in hierarchical instruction following, improving compliance and utility across various large language models.

Contribution

It formulates hierarchical instruction following as a Constrained Markov Decision Process and employs primal-dual RL to explicitly enforce prompt compliance.

Findings

Significantly improves system prompt compliance.

Enhances user utility across multiple model architectures.

Promotes attention shift toward long-range system tokens.

Abstract

Hierarchical Instruction Following (HIF) refers to the problem of prompting large language models with a priority-ordered stack of instructions. Standard methods like RLHF and DPO typically fail in this problem since they mainly optimize for a single objective, failing to explicitly enforce system prompt compliance. Meanwhile, supervised fine-tuning relies on mimicking filtered, compliant data, which fails to establish the priority asymmetry at the algorithmic level. In this paper, we introduce \textsc{HIPO}, a novel alignment framework that formulates HIF as a Constrained Markov Decision Process. \textsc{HIPO} elevates system prompts from mere input context to strict algorithmic boundaries. Using a primal-dual safe reinforcement learning approach, the algorithm dynamically enforces system prompt compliance as an explicit constraint, maximizing user utility strictly within this feasible region. Extensive evaluations across diverse model architectures (e.g., Qwen, Phi, Llama) demonstrate that \textsc{HIPO} significantly improves both system compliance and user utility. Furthermore, mechanistic analysis reveals that this constrained optimization autonomously drives the model to shift its attention toward long-range system tokens, providing a principled foundation for reliable LLM deployment in complex workflows.