AgentDropoutV2: Optimizing Information Flow in Multi-Agent Systems via Test-Time Rectify-or-Reject Pruning

2026-02-26

AI Summary

This work addresses the challenge of cascading errors in multi-agent systems, where misinformation from individual agents can severely degrade collective reasoning performance. To mitigate this, the authors propose a test-time framework that dynamically optimizes information flow by intercepting agent outputs and iteratively correcting them using a retrieval-augmented corrector informed by failure-mode priors and context-aware mechanisms. Irreparably erroneous outputs are pruned to prevent error propagation. Notably, the approach requires no retraining and enables adaptive error identification and mitigation. Evaluated across multiple mathematical reasoning benchmarks, the method achieves an average accuracy improvement of 6.3 percentage points, demonstrating strong generalization, adaptability to varying task difficulty, and robustness against diverse error patterns.

Technology Category

• Multiagent Systems: Adversarial Agents
• Knowledge Representation and Reasoning: Reasoning with Beliefs
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

While Multi-Agent Systems (MAS) excel in complex reasoning, they suffer from the cascading impact of erroneous information generated by individual participants. Current solutions often resort to rigid structural engineering or expensive fine-tuning, limiting their deployability and adaptability. We propose AgentDropoutV2, a test-time rectify-or-reject pruning framework designed to dynamically optimize MAS information flow without retraining. Our approach acts as an active firewall, intercepting agent outputs and employing a retrieval-augmented rectifier to iteratively correct errors based on a failure-driven indicator pool. This mechanism allows for the precise identification of potential errors using distilled failure patterns as prior knowledge. Irreparable outputs are subsequently pruned to prevent error propagation, while a fallback strategy preserves system integrity. Empirical results on extensive math benchmarks show that AgentDropoutV2 significantly boosts the MAS's task performance, achieving an average accuracy gain of 6.3 percentage points on math benchmarks. Furthermore, the system exhibits robust generalization and adaptivity, dynamically modulating rectification efforts based on task difficulty while leveraging context-aware indicators to resolve a wide spectrum of error patterns. Our code and dataset are released at https://github.com/TonySY2/AgentDropoutV2.

Problem

• Multi-Agent Systems
• error propagation
• information flow
• cascading errors
• test-time optimization

Innovation

• test-time pruning
• multi-agent systems
• error rectification
• retrieval-augmented correction
• failure-driven indicators