LLM Multi-Agent Systems Are Vulnerable to Herding and Runaway Cascades
Resumo indisponível. Consulte o paper original.
PAPER / ARXIV:2610.12988
Junzhou Su, Yichen Luo, Yebo Feng, Teng Li, Jiahua Xu
RESUMO
Cryptocurrency swing trading spans mismatched timescales: positions remain active while market and event risk can change within minutes. However, many predictive and LLM-based trading systems assess risk after strategy generation or run both functions at the same cadence, leaving proposals unconditioned by current risk. We propose SwingGuard, a dual-timescale, long-only multi-agent framework for risk-conditioned strategy generation. At the strategy timescale, specialized analysts synthesize technical, sentiment, on-chain, and macroeconomic evidence through bullish-bearish deliberation, combined by a Portfolio Architect into a target position. At the governance timescale, a five-minute Risk Controller performs conditional diagnosis, audits the proposal, and constrains execution until the next strategy revision. On Bitcoin and Ether, the exposure-capped joint portfolio achieves a Sharpe ratio of 2.98, versus 2.79 for an otherwise identical Post-hoc Risk variant, with structured traces linking risk evidence to strategy constraints for auditability.
NO MESMO MAPA
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.