Agent Bazaar is a multi-agent simulation framework that identifies economic failure modes in LLM agents, proposes stabilizing harnesses, and shows that targeted RL training can produce a 9B model with superior economic alignment compared to frontier models.
Training language models to follow instructions with human feedback.Advances in neural information processing systems, 35:27730–27744
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Agent Bazaar: Enabling Economic Alignment in Multi-Agent Marketplaces
Agent Bazaar is a multi-agent simulation framework that identifies economic failure modes in LLM agents, proposes stabilizing harnesses, and shows that targeted RL training can produce a 9B model with superior economic alignment compared to frontier models.