StateGuard: Analytical-State Management with Validity-Aware Intervention for Long-Horizon Data Agents

2026-09-28

AI Summary

This study addresses the issue of implicit states in long-horizon data analysis, where outdated information can erroneously propagate along dependency chains. To mitigate this, we propose a management framework based on externalized state graphs that explicitly models analytical progress. By integrating counterfactual supervised fine-tuning with validity-guided policy optimization, the framework enables systematic tracking, verification, and automatic repair of states. Experimental results demonstrate that our approach significantly enhances data agent performance across multiple benchmarks by effectively suppressing downstream error cascades, thereby improving reliability in complex, multi-stage tasks.

Abstract

LLM-based agents have shown strong capabilities in automated data analysis and are increasingly moving toward long-horizon, multi-stage analytical workflows. However, as the analytical process evolves, constraints, variables, and conclusions remain implicitly embedded in interaction histories, making it difficult for agents to track which analytical artifacts remain valid over increasingly long horizons and changing dependencies. Consequently, stale artifacts may be silently inherited, propagating errors to downstream stages. To address this challenge, we propose StateGuard, an analytical-state validity management framework for long-horizon data agents. StateGuard externalizes evolving analytical progress into a state graph containing constraints, versioned variables, intermediate conclusions, and cross-state relations, treating each state as an executable, verifiable, and traceable object rather than textual memory alone. StateGuard maintains state validity through evidence-grounded verification and hierarchical intervention. To equip StateGuard with these capabilities, we first introduce Manager-Oriented Counterfactual Supervision, which constructs 3K state-centric trajectories through counterfactual runtime synthesis to fine-tune StateGuard for state maintenance, verification, and repair. We then apply Validity-Guided Policy Optimization, using runtime validity evidence to provide fine-grained learning signals for protocol correctness, state grounding, and intervention quality. Experiments on three diverse long-horizon data-analysis benchmarks show that StateGuard consistently improves data-agent performance while reducing dependency-induced downstream error propagation, demonstrating the advantages of explicit analytical-state management for reliable long-horizon data analysis.

Problem

• long-horizon data agents
• analytical-state management
• error propagation
• stale artifacts
• LLM-based agents

Innovation

• Analytical-State Management
• State Graph
• Counterfactual Supervision
• Validity-Guided Policy Optimization
• Long-Horizon Data Agents