May 16, 2026 · cs.MAJ/K move · Enter open · S save
Weiyi Kong, Shiyang Lai, Jinghua Piao, James Evans
Department of Electrical Computer Engineering, University of Toronto; Toronto, M5S 3G4, Canada · Department of Sociology, University of Chicago; Chicago, 60637, USA · 3Knowledge Lab, University of Chicago; Chicago, 60637, USA · Department of Electronic Engineering, Tsinghua University; Beijing, 100084, China · 5Google, Paradigms of Intelligence Team; Mountain View, 94043, USA
Whether machines can originate novel content has been debated for nearly two centuries, from Lovelace's assertion that no engine can "originate anything" to Turing's question of whether a machine can amplify ideas brought in from outside. Systems of multiple interacting LLMs, increasingly deployed for autonomous generation, reopen this question empirically. Here we show that such systems, operating in inference-only setups, exhibit semantic collapse: systematic convergence in semantic representations despite apparent lexical variation. Across model families, extended simulations of 200 to 1,000 rounds, the pattern remains consistent. Thirteen intervention strategies, spanning decoding parameters, prompt design, agent composition, activation engineering, and reinforcement learning, fail to restore semantic diversity. Mechanistic analyses suggest that semantic collapse is not explained by alignment or conformity biases, but is consistent with intrinsic properties of autoregressive generation. Our results point to persistent constraints on the ability of multi-LLM systems to sustain open-ended exploration in closed-loop settings.