Introduces Contagion Networks framework and measures preference propagation in 3-agent LLM setups, finding architectural priors dominate prompts, topology affects spread, and larger committees reduce contagion by ~69%.
Zhang et al.Self-Consuming Performative Loop.arXiv:2601.05184, 2025
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
In multi-model self-consuming loops, cross-influences can dampen or invert the alignment benefits of human curation unlike in isolated single-model cases.
citing papers explorer
-
Contagion Networks: Evaluator Preference Propagation in Multi-Agent LLM Systems
Introduces Contagion Networks framework and measures preference propagation in 3-agent LLM setups, finding architectural priors dominate prompts, topology affects spread, and larger committees reduce contagion by ~69%.
-
When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop
In multi-model self-consuming loops, cross-influences can dampen or invert the alignment benefits of human curation unlike in isolated single-model cases.