Tag
This paper introduces Boundary_Sync, a protocol measuring representational coupling in multi-agent LLM systems. Experiments with GPT-4o show that communication causes significant homogenization in both text and image modalities, but group size moderates the direction, with larger groups leading to homogenization and smaller groups to diversification.
This paper presents a comprehensive 33-class taxonomy of recurrent distortion patterns (heuristic parasites) in LLM outputs, along with operational definitions, recognition criteria, and a reproducible measurement protocol (PPE) for quantifying behavioral degradation across conversations.