E-Ledger improves safe tool use in complex workflows with persistent tracking
Safe, Persistent, and Evolving Agent Harness for Understanding Partially Observable Worlds
Artificial Intelligence
Summary
Completing tasks in complex work environments is tricky because important rules and hidden effects are often unknown or unclear. The authors created E-Ledger, a system that checks every action against company policies and keeps a detailed record of what’s happened to ensure safety and consistency. They also made WorldAbduct, which helps E-Ledger learn hidden rules by analyzing past actions and checking hypotheses through special interactions. Together, these tools make completing long, complicated tasks safer and more reliable. Testing showed better results across multiple language models and different environments.
What this means in practice
- •For enterprise workflow teams: Support safe and compliant execution of complex workflows that involve multiple tools and hidden business rules.
- •For scientific simulation developers: Improve safety and rule discovery in scientific task simulations by managing uncertain and evolving environmental rules.
Authors
Yisen Gao, Yue Guo, Qing Zong, Yiwen Guo, Yangqiu Song
Abstract
Large language model agents can invoke tools fluently, but enterprise workflows demand more than selecting the right tools: actions must strictly comply with organizational policies, tool feedback often conceals hidden side effects under partial observability, and long-horizon tasks require persistent state tracking across multiple records. To address these challenges, we introduce E-Ledger, a multi-agent harness for safe and persistent execution. E-Ledger employs a code approval layer that checks every proposed action against policy before execution, and maintains a world ledger of verified hidden rules alongside evidence-backed dynamic state. Because hidden rules are typically unknown a priori, we further propose WorldAbduct, an abductive, world-model-driven harness evolution framework. WorldAbduct diagnoses execution trajectories across four complementary views (state consistency, world-observation gap, policy-gate correctness, and goal judgment) to hypothesize latent rules, and verifies them through targeted abductive interactions before integrating them into the ledger. On the enterprise benchmark World of Workflows, E-Ledger with WorldAbduct improves safe task completion across four LLM backbones, outperforming the strongest evolution baseline by 5--15 percentage points. Experiments in ScienceWorld and DiscoveryWorld further show that abductive harness evolution carries over to scientific environments. Our code is available at https://github.com/HKUST-KnowComp/E-LEDGER-WorldAbduct.