LLM agent committees exhibit representational collapse with mean cosine similarity of 0.888, and diversity-aware consensus reaches 87% accuracy on GSM8K versus 84% for self-consistency at lower cost.
Exploring system 1 and 2 communication for latent reasoning in LLMs
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Representational Collapse in Multi-Agent LLM Committees: Measurement and Diversity-Aware Consensus
LLM agent committees exhibit representational collapse with mean cosine similarity of 0.888, and diversity-aware consensus reaches 87% accuracy on GSM8K versus 84% for self-consistency at lower cost.