Market Regime Council for Dynamic Credit Assignment in Multi-Agent LLM Decision Systems
The article presents a new cooperative multi-agent decision system called Market Regime Council (MRC) for portfolio management. MRC addresses issues of credit assignment among agents and improves transparency in decision-making. The system has shown significant performance improvements in trading across various crypto assets.
- ▪MRC computes exact Shapley credits for online agent weighting across multiple outputs.
- ▪The system achieved a Sharpe ratio of 1.51 and a cumulative return of 440.1% over 1,037 trading days.
- ▪Ablation results indicate that performance gains are due to Shapley-weighted integration rather than individual stages.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.24490 (cs) [Submitted on 23 May 2026] Title:Market Regime Council for Dynamic Credit Assignment in Multi-Agent LLM Decision Systems Authors:Yunhua Pei, Zerui Ge, Jin Zheng, John Cartlidge View a PDF of the paper titled Market Regime Council for Dynamic Credit Assignment in Multi-Agent LLM Decision Systems, by Yunhua Pei and 3 other authors View PDF HTML (experimental) Abstract:Multi-agent LLM decision systems for portfolio management still lack a principled way to assign credit across specialist agents, remain vulnerable to cold-start dominance under regime shifts, and offer limited transparency into how final allocations are formed.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
