Framework improves repair of multi-agent system task failures
Dittany: Contribution-Guided Repair Framework for MAS Applications
Software Engineering
Summary
Multi-agent systems sometimes struggle to finish tasks completely, even though parts of their work are correct. The authors present a method called Dittany that helps find what work is unfinished and what results can still be used. Dittany then repairs the system by either fixing small parts or reconsidering larger dependencies, checking its fixes before continuing. In testing, it caught and fixed failures better than existing methods while using fewer computing resources.
What this means in practice
- •For software engineers: Fix incomplete or incorrect task executions in multi-agent AI workflows more accurately and efficiently at runtime.
- •For ai system integrators: Enhance robustness in multi-agent collaboration platforms by identifying remaining work and guiding appropriate repairs automatically.
Authors
Hanqi Sun, Yuchen Shao, Hanyi Si, Beibei Xu, Lisheng Wang, Chengcheng Wan
Abstract
LLM-based multi-agent system (MAS) applications can produce useful intermediate results while leaving task requirements unsatisfied. Repairing such executions requires identifying the work still needed and the completed results that can support it. We present \tool, a contribution-guided runtime repair framework for MAS applications. \tool derives expected contributions and acceptance conditions from task requirements and method specifications, and connects their dependencies in a contribution graph. Runtime checks locate unmet conditions and establish which results remain usable. These results guide action-level correction when available inputs suffice, or graph-level recovery when prerequisites or unfinished work need revision. Both repair scopes retain validated contributions and check repaired outputs before subsequent work uses them. We evaluate \tool on 30 applications across six MAS frameworks and 9,000 tasks. It detects runtime failures with 88.28\% precision and 79.18\% recall, and achieves 84.76\% task accuracy, improving over Native execution by 26.74 percentage points and the strongest of four recovery baselines by 21.28 points. Estimated deployment token consumption is 17.92\% above Native, but 24.19\% below Reflexion and 13.05\% below Iterative Rollback. Ablation results show that combining the two repair scopes achieves higher accuracy and lower token consumption than graph-level repair alone.