PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM Agents

Hugging Face Daily Papers Papers
llm-agents policy-compliance workflow-guidance safety-verification academic-research llm-safety

Summary

PolicyGuide compiles domain policies into workflow graphs and uses a proactive verifier to guide LLM agents through multi-step procedures, improving policy compliance across various benchmarks.

Customer-service LLM agents must follow organizational policy when acting on a user's behalf. Compliance failures arise from either forbidden actions, such as granting an ineligible change, or omitted procedural requirements, such as identification or confirmation. Runtime safeguards can intervene on risky actions, but action-local checks do not guide an agent through a multi-step procedure. Workflow-following systems support prescribed process execution, but primarily target workflow completion rather than safeguarding agent behavior. PolicyGuide instead compiles each domain policy into a workflow graph and invokes a proactive verifier at user-turn boundaries. From persisted graph state, the verifier reconciles open requests and returns step-specific remediation along a policy-compliant path. Across the τ^2-bench airline, retail, and telecom domains with a GPT-5.4 agent and verifier, PolicyGuide raises mean Pass^4 from 0.42 to 0.62, with the largest gain on telecom (0.19 to 0.61), the most workflow-structured domain. The same workflows transfer to Claude Sonnet 4.6 and Gemini 2.5 Pro agents. Complementary evaluations find the lowest observed attack-success rate under adversarial users and the strongest procedural compliance in an author-designed workflow-level validation.
Original Article
View Cached Full Text

Cached at: 08/21/26, 04:10 AM

Paper page - PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM Agents

Source: https://huggingface.co/papers/2608.19861

Abstract

Customer-serviceLLMagentsmustfolloworganizationalpolicywhenactingonauser’sbehalf.Compliancefailuresarisefromeitherforbiddenactions,suchasgrantinganineligiblechange,oromittedproceduralrequirements,suchasidentificationorconfirmation.Runtimesafeguardscaninterveneonriskyactions,butaction-localchecksdonotguideanagentthroughamulti-stepprocedure.Workflow-followingsystemssupportprescribedprocessexecution,butprimarilytargetworkflowcompletionratherthansafeguardingagentbehavior.PolicyGuideinsteadcompileseachdomainpolicyintoaworkflowgraphandinvokesaproactiveverifieratuser-turnboundaries.Frompersistedgraphstate,theverifierreconcilesopenrequestsandreturnsstep-specificremediationalongapolicy-compliantpath.Acrosstheτ^2-benchairline,retail,andtelecomdomainswithaGPT-5.4agentandverifier,PolicyGuideraisesmeanPass^4from0.42to0.62,withthelargestgainontelecom(0.19to0.61),themostworkflow-structureddomain.ThesameworkflowstransfertoClaudeSonnet4.6andGemini2.5Proagents.Complementaryevaluationsfindthelowestobservedattack-successrateunderadversarialusersandthestrongestproceduralcomplianceinanauthor-designedworkflow-levelvalidation.

View arXiv pageView PDFProject pageGitHub0Add to collection

Get this paper in your agent:

hf papers read 2608\.19861

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.19861 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.19861 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.19861 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

PolicyBank: Evolving Policy Understanding for LLM Agents

arXiv cs.CL

PolicyBank proposes a memory mechanism that enables LLM agents to autonomously refine their understanding of organizational policies through iterative interaction and corrective feedback, closing specification gaps that cause systematic behavioral divergence from true requirements. The work introduces a systematic testbed and demonstrates PolicyBank can close up to 82% of policy-gap alignment failures, significantly outperforming existing memory mechanisms.

Learning Agentic Policy from Action Guidance

arXiv cs.CL

The paper proposes ActGuide-RL, a method for training agentic policies in LLMs by using human action data as guidance to overcome exploration barriers in reinforcement learning without extensive supervised fine-tuning.

Governance by Construction for Generalist Agents

arXiv cs.AI

This paper presents CUGA's policy system, a modular policy-as-code layer that enforces governance at multiple checkpoints in LLM agent execution, enabling predictable and auditable behavior without model fine-tuning.

Autoformalization of Agent Instructions into Policy-as-Code

arXiv cs.AI

This paper presents an autoformalization pipeline that translates agent prompts, MCP tool descriptions, and natural language policy documents into formally verified policies using an LLM-based generator-critic loop, achieving better coverage than hand-coded enforcement on MedAgentBench.