ProMAS: A Proactive Approach to Error Analysis in Multi-Agent Systems

The integration of Large Language Models (LLMs) into Multi-Agent Systems (MAS) has enabled the solution of complex tasks through collaborative reasoning. However, this collective intelligence is fragile, as a single logical fallacy can rapidly propagate and lead to system-wide failure.

Current research relies primarily on post-hoc failure analysis, hindering real-time intervention. To address this, PROMAS, a proactive framework utilizing Markov transitions for predictive error analysis, has been proposed.

PROMAS extracts Causal Delta Features to capture semantic displacement, mapping them to a quantized Vector Markov Space to model reasoning as probabilistic transitions. By integrating a Proactive Prediction Head with Jump Detection, the method localizes errors via risk acceleration rather than static thresholds.

On the Who&When benchmark, PROMAS achieves 22.97% step-level accuracy while processing only 27% of reasoning logs. This performance rivals reactive monitors like MASC while reducing data overhead by 73%. Although this strategy entails an accuracy trade-off compared to post-hoc methods, it significantly improves intervention latency, balancing diagnostic precision with the real-time demands of autonomous reasoning.