EGL-SCA: Structural Credit Assignment for Co-Evolving Instructions and Tools in Graph Reasoning Agents

2026-05-11

AI Summary

This work addresses the challenge that graph reasoning agents must jointly handle natural language-to-structured-graph reconstruction, tool adequacy assessment, and structural correctness verification, yet existing approaches optimize instructions or tools in isolation, hindering precise failure attribution. To overcome this, the paper introduces the EGL-SCA framework, which pioneers a structural credit assignment mechanism that maps execution trace evidence into conditional update signals, enabling co-evolution of instruction policies and tool program spaces. Integrated with hierarchical task-family training distributions and a Pareto retention strategy, the framework balances success rate, generalization, and simplicity. Evaluated across four graph reasoning benchmarks, it achieves an average success rate of 92.0%, substantially outperforming prompt-only and fixed-toolbox baselines.

Abstract

Graph reasoning agents operating from natural-language inputs must solve a coupled problem: they must reconstruct a structured graph instance from text, decide whether existing computational assets are sufficient, interact with tools under a strict execution protocol, and satisfy an external verifier that checks structured correctness rather than textual plausibility. Existing approaches usually improve either the instruction side or the tool side in isolation, which leaves unclear what should be updated after failure. We propose EGL-SCA, a verifier-centric dual-space framework that models a graph reasoning agent using two collaborative components: an instruction-side policy space for reasoning strategies, and a tool-side program space for executable algorithmic tools. Our central mechanism is structural credit assignment, which maps trajectory evidence to conditional updates, precisely routing failures to either prompt optimization or tool synthesis and repair. To provide sufficient learning signals for dual-space adaptation, we introduce a training distribution stratified by task family, coupled with a Pareto-style retention strategy to balance success, generality, and parsimony. Experiments on four graph reasoning benchmarks show that EGL-SCA achieves a state-of-the-art 92.0\% average success rate. By effectively co-evolving instructions and tools, our framework significantly outperforms both pure-prompting and fixed-toolbox baselines.

Problem

• graph reasoning
• structural credit assignment
• instruction-tool co-evolution
• verifier-centric learning
• natural-language to structured reasoning

Innovation

• structural credit assignment
• dual-space framework
• tool synthesis
• prompt optimization
• graph reasoning agents