NeFut Logo NeFut
Admin Login

[CS.AI] ADIAS: Automated Design of Interactive Agentic Systems

Published at: 2026-08-10 22:00 Last updated: 2026-08-11 02:05
#AI #Automated Design #Interactive Agentic Systems

ADIAS is a framework for automated design of interactive agentic systems. Existing methods are largely candidate-centric, which leaves the repair progress implicit and leads to inefficient repair targeting, slow consolidation of partial progress, and propagation of ineffective interventions across rounds. To address this, we formulate issue-centric agent optimization, where repair progress is carried forward as an explicit persistent issue state to guide optimization. ADIAS includes two mechanisms: persistent issue state and issue-guided optimization. The persistent issue state maintains stable issue identities, lifecycle status, supporting evidence, and intervention-outcome histories. Issue-guided optimization uses this state to jointly propose repair targets and revision directions for subsequent focused full-code modification. Across five interactive benchmarks, ADIAS outperforms the strongest baseline by 25.2% on average and achieves consistent gains across four backbone models. Controlled ablations further show that removing persistent issue state or replacing issue-centric revision with candidate-centric policies leads to performance drops of up to 40.7%. Blogger's Review: ADIAS framework achieves automated design of interactive agentic systems through issue-centric optimization, significantly improving performance and efficiency. The advantage of this method lies in its ability to explicitly track repair progress and problem states, enabling more effective optimization and revision.

Original Source: https://arxiv.org/abs/2608.06410

[h] Back to Home